Positive Selection on Multiple Antique Allelic Lineages of Transferrin in the Polyploid Carassius auratus
http://www.100md.com
分子生物学进展 2004年第7期
* State Key Laboratory of Freshwater Ecology and Biotechnology, Wuhan Center for Developmental Biology, Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, China
Modern Virology Research Center, College of Life Sciences, Wuhan University, Wuhan, China
E-mail: jfgui@ihb.ac.cn.
Abstract
Transferrin polymorphism has been studied in the polyploid Carassius auratus by cloning and sequence analysis of cDNAs from its three subspecies C. auratus gibelio, C. auratus auratus, and C. auratus cuvieri. DNA polymorphism of extremely high extent was shown for the transferrin gene by the 248 segregation sites among coding region sequences of its alleles. The deduced amino acid sequences of the transferrin alleles showed variable theoretical physicochemical parameters, which might constitute molecular basis for their electrophoretic heterogeneity. Positive selection was inferred by the replacement/synonymous ratios larger than 1 in partial allelic lineages which was subsequently confirmed by likelihood simulation under neutral or selection models. Furthermore, the correspondent sites to these selected codons were collectively located at two planes in the crystallographic structure of rabbit transferrin, which suggested that the rapid evolution of C. auratus transferrin might correlate to its adaptation to variable environmental elements such as oxygen pressure. The minimal 26 recombination events were detected among coding sequences of C. auratus transferrin, with partial mosaic sequences and breakpoints identified by identity scanning and information site analyses. Phylogenetic analyses revealed multiple antique allelic lineages of transferrin, which was estimated to diverge fifteen to twenty MYA. All these features strongly suggested the role of balancing selection in long persistence of high transferrin polymorphism in C. auratus. Furthermore, owing to its particular evolutionary backgrounds, the silver crucian carp might possess a distinctive balancing selection mechanism.
Key Words: transferrin ? polymorphism ? positive selection ? balancing selection ? allele ? recombination ? Carassius auratus
Introduction
Transferrin is a protein involved in iron storage and transport of vertebrates. Abundant polymorphism was observed in electrophoretic phenotypes of transferrin in many species, and its zymograms were often used as genetic markers for studying of population genetics (Kirpichnikov 1981). Meanwhile, many efforts have been made to elucidate the genetic basis of this polymorphism. In human, transferrin variants of different electrophoretic migrations have been successively identified, and most of them exhibit similar physiochemical properties and functions despite the occasional replacements among their amino acid (aa) sequences (Welch and Langmead 1990). In horse and cattle, transferrin polymorphism was studied at the gene level, and a series of single nucleotide polymorphism sites and aa substitutions respectively were identified. This had partially interpreted the heterogeneous phenotypes of different transferrin variants (Carpenter and Broad 1993; Laurent and Rodellar 2001). Furthermore, posttranslational glycosylation was also characterized in some species responsible for electrophoretic variation of transferrin (Williams 1968).
In studying transferrin polymorphism, attempts have been made to elucidate its evolutionary mechanism. The often high level of polymorphism in transferrin was thought to be maintained by a preferable selection for heterozygotes based on the observation of heterozygous advantage or high heterozygosity within populations (Frelinger 1972; Kirpichnikov 1981). Ford (2000, 2001) studied molecular evolution of transferrin in the salmonids. Positive selection was confirmed to play significant roles in diversification of transferrin molecules among salmonid species, and localization of the selected sites suggested the competition for iron from pathogenic bacterial as the evolutionary drive (Ford 2001). Recently, evidences for positive selection of tranferrin were further revealed at the intraspecific level in the brown trout, and recombination and gene conversion events responsible for diversification of transferrin alleles were also detected (Autunes et al. 2002).
In cyprinid fish, the species Carassius auratus and its subspecies have been widely recognized as model animals in evolutionary biology (Ohno 1970; Gui 1996; Zhou and Gui 2002). The common goldfish (C. auratus auratus) has a distribution throughout the Eurasia continent. It has 100± chromosomes (Yu et al. 1989) and multiple duplicated loci were identified from its genome (Wilson, Middleton, and Warr 1988; Risinger and Larhammar 1993), which inferred a probable tetraploidization event in its earlier evolution. The white crucian carp (C. auratus cuvieri), originating from Lake Biwa of Japan, also has 100± chromosomes (Gui et al. 1992). The silver crucian carp (C. auratus gibelio) has a distribution over a large territory over Europe, Russia, China and Japan. It has over 150 chromosomes and higher ploidy (Zhu and Jiang 1993; Zhou and Gui 2002; Wei et al. 2003; Yi et al. 2003). Two different reproductive modes were revealed in this subspecies, i.e., allogynogenetic reproduction and gonochoristic reproduction (Gui et al. 1993; Zhou, Wang, and Gui 2000; Zhou and Gui 2001). Transferrin polymorphism has been observed in the silver crucian carp, and three distinct transferrin phenotypes have been revealed among its three gynogenetic clones: A, D, and F (Yang et al. 2001). Three alleles were considered to genetically determine this polymorphism, and the heterozygotic transferrin genotypes apparently showed a clonal inheritance pattern. In a color variety of C.a. auratus, color crucian carp, the three transferrin variants A1, A2, and B1 have been identified, and substantial substitutions were observed among their alleles (Yang, Zhou, and Gui 2004).
In this study, we further cloned eight alleles of transferrin from the silver crucian carp and white crucian carp. Sequence variations among transferrin alleles of the polyploid C. auratus were studied comparatively, and positive selection and recombination were verified to play important roles in their diversification. Furthermore, multiple antique allelic lineages were identified, which inferred a balancing selection mechanism in maintenance of their polymorphism.
Materials and Methods
Specimens Collection and Electrophoretic Analysis of Transferrin Phenotypes
Three silver crucian carp individuals were selected for experiments, respectively representing the three silver crucian carp gynogenetic clones of A, D, and F (Yang et al. 2001). Two white crucian carp specimens were collected from an artificially maintained small population from the Duofu Farm. The common carp individual was identified to be from the Xingguo red common carp strain (Cyprinus carpio red variety). All of these six fish specimens had been raised in separate ponds of the Guanqiao Experimental Station, the Institute of Hydrobiology, the Chinese Academy of Sciences. The three silver crucian carp clones were identified by polyacrylamide gel electrophoresis (PAGE) analysis of their sera (Yang et al. 2001). Serum transferrin of the two white crucian carp individuals was also analyzed, and they showed different phenotypes.
Amplification of Transferrin cDNA by Polymerase Chain Reaction (PCR)
Total RNA was isolated from the fresh livers of the six fish specimens using Trizol reagent (GIBCO/BRL), and cloning of transferrin cDNA had begun from clone D of gynogenetic silver crucian carp. Degenerated primers Rt-5'/Rt-3', selected from the coding region of the transferrin gene in other vertebrates, were used for amplifying an original transferrin cDNA fragment (table 1). First strand cDNA was reverse-transcribed from 2 μg liver RNA in a 20-μl reaction mixture containing 200 u of M-MuLV reverse transcriptase (GIBCO BRL). One or two μl of this cDNA was used for amplification in a 50-μl reaction mixture containing 1 x buffer, 200 μM dNTP, 0.5 μM of each primer, and 2 units of Taq DNA polymerase (Biostar). The cycling profile was 1 cycle of 95°C for 5 min, 54°C for 2 min, and 72°C for 40 min followed by 30 cycles of 94°C for 1 min, 48°C for 1min, and 72°C for 3 min. Cloning and sequencing analysis confirmed this amplified product as the transferrin cDNA fragment. Furthermore, detection of two clones with the variable nucleotide sequences (designated as silver-A and silver-B) suggested the probable existence of multiple transferrin alleles (fig. 1). Gsp-5' and Gsp-3' primers were then selected from these sequences for amplification of the cDNA ends. First strand cDNA was initiated by primer Seq-3', recovered by use of NucleoTrap PCR Purification Kit (CLONTECH), and then added with a poly(A)+ tail using terminal transferase (Promega). Four μl of this cDNA was used for amplification of the 5' transferrin cDNA ends in a 50-μl cocktail containing 0.1 μM Adap-(T)17 and 0.5 μM Gsp-5'. 0.5 μM of Adap primer was added after the initial cycle of 95°C for 5 min, 43°C for 2 min, and 72°C for 40 min. This was followed by 30 cycles of 94°C for 1 min, 58°C for 1 min, and 72°C for 1 min. Amplification of the 3' cDNA ends was performed as described for Rt-5'/Rt-3' amplification except that Adap was used for the downstream PCR primer.
Table 1 Primers Used for the Amplification of C. auratus Transferrin cDNA.
FIG. 1. Amplification of transferrin cDNA sequences from D clone of gynogenetic silver crucian carp. (top section) Primary structure of crucian carp transferrin cDNAs, with shaded box and lines displaying the open reading frame and the untranslated regions, respectively. (bottom section) Three approximately complete transferrin cDNA sequences were amplified. First, a transferrin cDNA fragment was amplified by PCR with degenerated primers Rt-5'/Rt-3', with alleles silver-A and silver-B identified. Then, 5' and 3' cDNA ends were PCR amplified, but a complete cDNA sequence could not be presented for the variable Rt-5'/Rt-3' sequences. Thus, a nearly complete transferrin cDNA sequence had to be amplified by gene-specific For/Rev, by which the third transferrin allele, silver-C, was identified. Allele-specific PCR were then employed to amplify cDNA ends of silver-A or silver-B, and their complete cDNA sequences were thus obtained. The pattern of strand indicates cDNA sequences of different transferrin alleles: black, of silver-A; half-transparent, of silver-B; and transparent, of silver-C. The numbers of sequenced clones are listed on the PCR products. An asterisk (*) indicates that the cDNA fragment was from the silver-A' transferrin allele, with sequence identical to that of silver-A in this region
Unexpectedly, no sequence variation was detected among clones from the 5' or 3' transferrin cDNA ends; thus, a complete transferrin cDNA sequence could not be presented by splicing of the three available PCR products (fig. 1). The gene-specific primer set For/Rev was selected from the cDNA ends to amplify a complete transferrin cDNA sequence directly, by which the third transferrin allele, silver-C, was identified. Afterwards, primers specifically annealing to cDNA sequences of certain alleles such as Var-5'(A) and Var-3'(A) were employed to amplify their 5' and 3' ends (table 1); ultimately, their integrate cDNA sequences were obtained.
With similar procedures, variable transferrin cDNA sequences were amplified from liver RNA of silver crucian carp clone A and clone F; two other alleles, silver-D and silver-E, were then identified from these two carp clones, respectively. With respect to white crucian carp, cloning of transferrin cDNA sequence was accomplished by sequencing of four clones of For/Rev PCR products, with two from the two individuals each, by which three transferrin alleles, white-A, white-B and white-C, were detected. Detection of transferrin alleles from the six C. auratus individuals is listed in table 2.
Table 2 Detection of Transferrin Alleles Among C. auratus Individuals by RNA PCR.
Cloning and Sequence Determination of PCR Products
PCR products were isolated by agarose gel electrophoresis using NucleoTrap Gel Extraction Kit (CLONTECH). Recovered DNA fragments were directly cloned into pGEM-T vector as the manufacturer described (Promega). One to three clones were subjected to sequence determination for each of the PCR products, except that eight ones were sequenced for Rt-5'/Rt-3' product to detect more plausible transferrin alleles (refer to fig. 1 for the condition in clone D). Sequence analysis was programmed using dRhodamine terminator cycle sequencing Kit on ABI PRISMTM 310 Genetic Analyzer (Perkin Elmer). Translation of aa sequences, predication of signal peptide, and computation of isoelectric point and molecular weight (pI/Mw) were accomplished by use of software at ExPASy Molecular Biology Server (http://expasy.pku.edu.cn).
Estimation of dn/ds Ratios and Selected Sites
Besides the eight transferrin coding sequences presented here (table 3), two cDNA sequences had been described for color crucian carp previously (GenBank accession numbers AF518746 and AF518747; Yang, Zhou, and Gui 2004). All these sequences were aligned using the ClustalX 1.8 program (Thompson et al. 1997) with default alignment parameters (gap opening penalty 10.00; gap extension penalty 0.20; sequences >30% diverged delayed; DNA transition weight penalty 0.50; with negative matrix off). The obtained sequence alignments were 1,965 base pairs long, without complete 5' ends and any gaps.
Table 3 Theoretical Physiochemical Properties of Transferrin Variants from C. auratus.
For analysis of the plausible selective drives in the evolution of transferrin, a phylogenetic tree was constructed from the aligned DNA sequences using the maximum-likelihood method in the DNAML program of PHYLIP3.5 package (Felsenstein 1993; see fig. 3 for detailed parameters). Furthermore, the PAML computer package (Yang 1997) was used to examine the possible evolutionary models of the transferrin of C. auratus. Four codon-based models—single-ratio, free-ratio, neutral, and selection—were successively applied on different molecular data (Nielsen and Yang 1998; Yang 1998). Of the many parameters involved, "model for codons" and "NSsites" were especially considered for their crucial effects on selection of evolutionary models. Following were their settings for the four given models in order: 0, 0; 1, 0; 0, 1; 0, 2. The fitness of different models was examined by considering the 2 distribution of 2l, twice the log likelihood difference between any of the two models, with the model of likelihood value significantly larger than the preferable one. In the selection model, codons subject to selection were identified by their high posterior probability in the empirical Bayes's approach (Nielsen and Yang 1998). Furthermore, the sites had to show a replacement substitution rate (dn) larger than the silent substitution rate (ds) for confirmation of their positive selection. The software Cn3d3.0 was available from NCBI (www.ncbi.nlm.nih.gov) for the illustration and disposal of the three-dimensional structure of rabbit transferrin.
FIG. 3. Maximum-likelihood phylogeny based on transferrin coding sequences of C. auratus. The tree topology was estimated using the DNAML program in the PHYLIP package (Felsenstein 1993). Numbers next to the branches are the dn/ds ratios estimated under the free-ratio model of PAML package (Yang 1998)
Recombination Analysis
After removal of the 11 selected codons (fig. 2), population recombination rates and the minimal set of recombination intervals (Hudson and Kaplan 1985) were estimated for transferrin sequences C. auratus (with 1,932 nucleotides left) using the program SITES version 1.1 (Hey and Wakeley 1997). Because incomplete reproduction isolation and apparent migration happened in the divergence of the three C. auratus subspecies, transferrin sequences from all of them and uniquely from the gynogenetic silver crucian carp were subject to calculation respectively for comparison.
FIG. 2. Alignment of putative aa sequences of transferrin from C. auratus and common carp (GenBank accession no. AF457152; Yang, Zhou, and Gui 2004). Transferrin alleles of C. auratus and their encoded proteins are abbreviated: sil_A for silver-A and sil-A for silver-A, respectively, etc. Dots (.) represent aa residues identical to those of transferrin variant A from the silver crucian carp. Gaps (-) are introduced to optimize identity. Conserved regions involved in the iron ligand and anion binding of transferrin are shaded. Asterisks (*) represent the iron or anion ligands. The predicted signal peptide sequences of transferrins are shown by underlined letters. The signal "#" marks the potential residues for glycosylation in white-A and white-B. Triangles indicate the sites subject to positive selection. The program ClustalX 1.8 was used for sequence comparison and alignment (Thompson et al. 1997)
The recombination intervals and breakpoints were then determined for transferrin alleles by use of the software SIMPLOTS2.5 (Robertson et al. 1995). The plausible mosaic sequences were first identified from coding sequences of the transferrin alleles by querying each one to the others using the Plot and BootScan functions. Four sequences were sorted out to form a group including the mosaic allele, the two plausible parental alleles, and a background allele. Informative sites were identified for these sequences with each one supporting one possible phylogeny. Locations of the breakpoints were then determined by their maximizing the 2 value among the phylogeny clusters. P-values for the resultant divisions of sites were calculated using Fisher's exact test.
Evolutionary Rate Tests and Phylogenetic Analysis
Recombination and selection affect the molecular clock (Arbogast et al. 2002); thus, divergence time estimation requires consideration of the interference of recombination and selection. Becuase recombination was detected primarily through the intermediate region of transferrin coding sequence, the homogenous regions of the 5' and 3' ends were selected for the molecular clock test and phylogeny reconstruction using the LINTREE program (Takezaki, Rzhetsky, and Nei 1995). A number of transferrin sequences were cloned from salmonids (refer to Ford 2001 for index), and divergence of these species might provide an ideal calibration point for molecular clock calculation. Nine of their transferrin sequences are cited here (with their GenBank accession numbers listed in parentheses): Atlantic salmon (L20313; Kvingedal, Rorvik, and Alestrom 1993), brown trout (D89091; Lee et al. 1998), lake trout (D89090; Lee et al. 1998), brook trout (D89089; Lee et al. 1998), Japanese Char (D89088; Lee et al. 1998), coho salmon (D89084; Lee et al. 1995), sockeye salmon (D89095; Lee et al. 1998), rainbow trout (D89083; Tange et al. 1997), amago salmon (D89086; Lee et al. 1998). Coding sequences of human (S95936; Hershberger et al. 1991) and chicken (X02009; Jeltsch and Chambon 1982) transferrin genes were also used as an outgroup. These 11 transferrin sequences were aligned with those of C. auratus using the default parameters of the ClustalX 1.8 program (Thompson et al. 1997). In the transferrin sites of C. auratus and salmonids subject to positive selection (Ford 2001), the 5' obscure nucleotide and all the gaps were removed from the sequence alignments when a 332-nucleotide alignment and a 530-nucleotide alignment were obtained for the middle region and 3' end of transferrin sequences.
Neighbor-joining trees were constructed by LINTREE, with evolutionary distance measured by Kimura's two-parameter distance (Kimura 1980). Credibility of the nodes was assessed by the interior branch test from the MEGA2.0 program (Kumar et al. 2001). Two cluster tests examined the difference between distances within each interior node, and the branch length test examined the deviation of the branch length between the tree root and a tip from the average length (Takezaki, Rzhetsky, and Nei 1995). Both of these differences were denoted by , and its deviation from zero (for two-cluster test) or from the average branch length (for branch length test) was valued by the two-tailed normal deviate test. The heterogeneous sequences were eliminated from trees until all the nodes or branches showed the constant evolutionary rate (P > 1%). A linearized tree was then constructed by re-estimation of the branch lengths.
Results
Transferrin cDNAs and Their Deduced Amino Acid Sequences
Besides two transferrin cDNA sequences from the color crucian carp (Yang, Zhou, and Gui 2004), eight novel transferrin cDNAs have been cloned in this study. They include five alleles (silver-A, -B, -C, -D, -E) from three gynogenetic clones (D, A, F) of the silver crucian carp (table 2) and three alleles (white-A, -B, -C) from two individuals of white crucian carp.
The transferrin cDNAs showed sequences of 2,228 (white-B) to 2,372 (silver-A) nucleotides. Although only two of these sequences (silver-A and silver-C) had poly (A)+ tails in the 3' ends, all of them contained complete open reading frames as well as their putative aa sequences; they are available from GenBank (table 3).
The transferrin alleles also encode variable lengths of protein sequences: 661 aa for silver-C; 666 aa for silver-E; 669 aa for silver-A and -B; 670 aa for white-B; and 671 aa for silver-D and white-A and -C (fig. 2; table 3). Meanwhile, they contain two lengths of signal peptide sequences: 15 aa for silver-C, -D, and -E and white-A and -B; and 21 aa for silver-A and -B and white-C. In contrast to the studies in many other species (Baldwin 1993; Lee et al. 1998), most transferrin variants of C. auratus have no potential N-glycosylation site and do not seem to be glycoproteins (with a single modification site found for white-A and -B as the exception). Plausible variations could also be found in these variants, and 34 and 37 cysteine residues were mainly observed in silver crucian carp and white crucian carp transferrins, respectively. After removal of the presumed signal peptides, Mw and pI values were calculated. As shown in table 3, the predicted Mw vary in a limited range: silver-D and white-A have the largest one of about 72.5 kDa; white-B has the second largest of 72.2 kDa; silver-A, silver-B, silver-E, and white-C are 71.9 kDa, approximately; and silver-C has the smallest of 71.6 kDa. The average Mw of 72 kDa is fundamentally consistent with the deduction from SDS-PAGE (Yang et al. 2001). With respect to pI, five transferrin variants show three theoretic values in silver crucain carp: silver-D and silver-E have the lowest one of 5.67; silver-A and silver-B have a higher one of 5.74; while silver-C has the highest one of 5.90. As to white crucian carp, three transferrin variants display similar pI values of about 5.90.
Sequence Comparison and Polymorphism
Polymorphism analysis showed abundant variation among the 10 transferrin cDNA sequences of C. auratus, with the segregation site number of 248. Their deduced aa sequences were then aligned together with that of common carp transferrin for comparison (fig. 2). In contrast to the few substitutions detected among transferrin variants in other species (Welch and Langmead 1990; Carpenter and Broad 1993), a large number of variations were observed throughout the over-600 aa transferrin sequences, which are mainly composed of replacements and deletions. A number of conserved residues had been identified for transferrins from different species (Anderson et al. 1989), including two tyrosines (Tyr), a histidine (His), and an aspartic acid (Asp) within each transferrin half. However, significant substitutions happened in the N-terminal domain of C. auratus transferrin: two Tyr residues for iron binding were replaced by a His-101 (for silver-C, color-B1, and white-A) and an Asp-199 (glutamine for color-A1), respectively, whereas the arginine (Arg) or lysine (Lys) for anion binding was substituted by a serine residue (Ser-129). Furthermore, substantial substitutions also happened in conserved regions around these residues, with a faster accumulating tendency in those of the N-terminal domain than in those of the C-terminal (12 vs. 3) of transferrin variants of C. auratus.
Sequence identity among transferrin alleles of C. auratus varies greatly (table 4). The highest sequence homology was observed between silver-C and color-B1 (99.7% for the cDNA sequences and 99.5% for the aa sequences), and the lowest one was observed between silver-A and silver-C (91.4% for the cDNA sequences and 84.3% for the aa sequences). The average identity for transferrin cDNA and aa sequences were 93.9% and 89.1%, respectively.
Table 4 Percentage Sequence Identity of cDNA Sequences and Amino Acid Sequences Between Transferrin Alleles of C. auratus.
Positive Selection and Identification of Selected Sites
For simulation with various evolutionary models, a phylogenetic tree was first constructed for transferrin alleles of C. auratus based on their coding sequences using the maximum-likelihood algorithm (Felsenstein 1993). As shown in figure 3, these alleles did not show a tendency to form a single cluster within a same subspecies. Instead, partial transferrin alleles from the silver crucian carp showed an apparently closer phylogenetic relationship with those from the color crucian carp, e.g., silver-C and color-B1.
The free-ratio model of the PAML program assumed different dn/ds values among branches of the evolutionary tree. Of the 17 branches of ml tree based on transferrin sequences of C. auratus, seven showed the dn/ds value greater than 1.0 (fig. 3). The branches of larger dn/ds values did not show a converging distribution in certain allelic lineage. The log likelihood value under the free-ratio model was l1 = –4854.90. Meanwhile, under the one-ratio model, which assumed common dn/ds numbers of different branches parameter for the entire tree, the calculation led to l0 = –4866.39. Twice the log likelihood difference of two models was 2l = (l1 – l0) = 22.98. No remarkably significant difference was detected between these two models by 2 distribution (0.10 < P < 0.25; table 5) with the free degree of 16. There seemed to be no detectable difference in the rates of adaptive evolution among allelic lineages of C. auratus transferrin.
Table 5 Results of Likelihood Models.
Further calculation under the positive-selection model and the neutral model was practiced (table 5). Twice the log likelihood difference between these two models was 53.60, and results of the 2 distribution test provided a much better fit for the former model than for the latter (df = 2, P < 0.005). Furthermore, approximately 14% of the total codons were subjected to probable positive selection during their evolutionary history by the calculation, with an average value of 4.29. The sites, subject to no or strong constraint, occupied 31% and 55%, respectively (table 5). All these data presented substantial evidence for the existence of positive selection in the evolution of transferrin of C. auratus.
Empirical Bayes's approach was further applied to detect codons subjected to positive selection (Nielsen and Yang 1998). Altogether, 11 sites were concluded as subject to selection in transferrin sequences, for their posterior probabilities greater than 0.95. These included the codons Ile70, Gly86, Ala97, Ala142, Ile151, Arg254, Pro297, Pro330, Arg358, Asn411, and Gly428 of silver-A in figure 2. None of these aa residues were distributed within the conserved regions for iron ligand or anion binding of transferrin.
The correspondent residues of these selected sites were then identified in the spatial structure of rabbit transferrin (fig. 4; Hall et al. 2002), with the peptides for binding of bacterial protein similarly illustrated for comparison (Retzer, Yu, and Schryvers 1999). As a result, 10 selected sites found their correlates in rabbit transferrin sequence with the Ile151 left for the gap in sequence alignment (data not shown). The selected sites generally fall around the peptides for binding of bacterial proteins with respect to the salmonid transferrins (Ford 2001). However, only three of the 11 selected sites from C. auratus transferrin were located within these peptides. A test of 2 distribution implied no definite correlation between the distribution of these codons and the location of these peptides (0.25 < P < 0.75; df = 1). However, five of the selected sites, Val-60, Asn-76, Lys-92, Ser-248, and Ala-299, were located convergent in a plane internal in N-lobe. This plane exactly paralleled with the interface between the two domains—N1 and N2—of N-lobe with the iron atom vertically over it. Furthermore, four of the five remnant selected codons, Cys-137, Pro-332, Leu-365, and Lys-418, lay in a long and narrow region on the back of linkage between the two lobes of rabbit transferrin, while the last, Asn-435, was located without direct connection with any of the other selected sites. A common attribute of these five sites was that they were distributed on the surface of the transferrin molecule.
FIG. 4. Tubular model of the diferric rabbit serum transferrin molecule (Hall et al. 2002). The molecule is specially oriented for a clear view of the selected sites. Upper right, upper left, lower left, and lower right approximately illustrate the N1, N2, C1, and C2 domains of transferrin, respectively. Peptides to be contacted by bacterial transferrin-binding proteins (Retzer, Yu, and Schryvers 1999) are shown by dark tubules. Five protein ligands involved in binding of iron atom or the anion in the N-lobe of transferrin are indicated by white balls. The sites correspondent to those subject to positive selection in C. aurtus transferrin are illustrated by dark balls, and the codon number in sequence alignment is marked. The codon Ile146 did not have homologous resiude in the rabbit transferrin
Population Recombination Rates and Mosaic Sequences
By the coalescent estimator of the SITE program (Hey and Wakeley 1997), the population recombination rates of transferrin alleles in the C. auratus and the silver crucian carp were 0.0807 and 0.0403, respectively. Since most of C. auratus and silver crucian carp were tetraploidy and hexaploidy, respectively (Zhu and Jiang 1993), these numbers should be split by two and three and the ultimate values should be 0.0404 and 0.0135. Furthermore, the minimal set of recombination intervals (Hudson and Kaplan 1985) for transferrin of the C. auratus and the silver crucian carp were 26 and 10, respectively.
Ten coding sequences of C. auratus transferrin alleles were subsequently queried to those of the others using SIMPLOT2.5. Recombination was evidently observed for four of these sequences, including silver-A, silver-D, silver-E, and color-A1. No mosaicism was found among coding sequences of any of the transferrin alleles of white crucian carp. Furthermore, transferrin alleles of white crucian carp appeared to have an earlier divergence from those of the other C. auratus populations (fig. 3 and fig. 6); thus, the allele white-A was used as the outgroup in the further informative site analyses for determination of breakpoints among mosaic sequences (fig. 5). Recombination could be easily identified for the coding sequence of silver-A by comparing its sequence identity with that of silver-B and silver-D (fig. 5A). In the 5' and 3' ends of these sequences, most of the informative site trees supported a closer relationship between silver-A and silver-B. In the intermediate region, however, silver-A and silver-D shared more informative sites. Two recombination events were deduced between silver-B and silver-D in producing silver-A, probably through nucleotides 561–691 and 1275–1305 (P < 0.001 for both results). Similarly, silver-D could be considered as a recombination product of silver-E and silver-A with the breakpoints between nucleotides 881 and 936, 1341 and 1347, respectively (fig. 5B; P < 0.001 for both results). The mosaic sequences of silver-E allele were from silver-D and color-A1, with the breakpoint between nucleotides 753 and 897 (fig. 5C; P < 0.005). Also, a recombination event most likely happened among nucleotides 233–540 between silver-E and silver-D to produce color-A1 (data not shown, 0.005 < P < 0.01). Based on all these analyses, recombination seemed not to happen between the two regions of 936 to 1275 and 1347 to 1932 of transferrin coding sequences, and they were used for molecular clock calculation in the following analyses.
FIG. 5. Similarity plot for the transferrin alleles (A) silver-A, (B) silver-D, (C) and silver-E, with a window size of 400 bp and a step size of 10 bp. The two allelic sequences putative as the recombination parents were selected as majority (50%) consensus sequences of the subject allele to display the mosaicism. Dotted vertical lines indicate breakpoints identified by maximization of 2 as described in Materials and Methods, with numbers of informative sites shared by the subject sequence and the reference sequences indicated below in the color to that reference. P-values were calculated by using Fisher's exact test. Four-member trees consistent with these sites are shown on the left
Molecular Clock Tests and Estimation of Divergence Time
Two neighbor-joining trees were generated based on the middle region and 3' end of transferrin coding sequences, and the former is presented in figure 6. As shown in figure 6, the transferrin sequences from C. auratus and salmonids formed one distinct cluster. The salmonid transferrins could be divided into three major clusters exactly correspondent to its three taxonomic genera Salmo, Salvelinus, and Oncorhynchus. By contrast, the transferrin alleles from each subspecies of C. auratus did not form one monophyletic group. Partial alleles were closely related to each other within each C. auratus subspecies, i.e., white-B and white-C, silver-A and silver-D. However, transferrin alleles among populations could also share a close phylogenetic relationship, i.e., silver-C and color-B1, silver-E and color-A1. These data suggest that the divergence between transferrin alleles of C. auratus might occur before the geographic differentiation of its subspecies.
All the interior nodes showed the fundamentally identical evolutionary rate in the two-cluster test (Takezaki, Rzhetsky, and Nei 1995). However, two of the silver crucian carp transferrin alleles, silver-A and silver-D, evolved significantly faster than other copies at the 1% level in the branch-length test. Elimination of these two sequences favored a constant evolutionary rate among nodes and branches in both tests, and the U statistic for the two-cluster test was 19.0 (0.25 < P < 0.75; df = 17). A linearized tree was then constructed (fig. 7). The nine transferrin sequences of salmonids again formed three distinct clusters correspondent to their three genera, the divergence of which (node b) was used as a calibration point for estimation of the divergence time of C. auratus transferrin (node a). The fossil records of salmonid genera were 14 to 17 MYA for Salvelinus (unavailable unpublished data) and 14 MYA for Oncorhynchus (S. Ralph and G. R. Smith, unpublished data). The average heights of nodes b and a were 0.0184 ± 0.0045 and 0.0274 ± 0.0047, respectively. By referring to the smaller salmonid divergence of 14 Myr, polymorphism of C. auratus transferrin might have existed for 20.8 ± 3.6 Myr.
FIG. 7. Linearized tree constructed under the assumption of a molecular clock for the middle-region transferrin coding sequences of human, chicken, salmonids, and C. auratus. The divergence of two salmonid genera, Salvelinus and Oncorhynchus (node b), was used as the calibration point to deduce divergence of transferrin alleles of C. auratus (node a). The scale at the bottom indicates nucleotide substitutions per site
The apparently different phylogenetic relationship between transferrin alleles of C. auratus was observed in the neighbor-joining tree from 3'-end transferrin coding sequences (data not shown). After removal of the four salmonid (Atlantic salmon, brown trout, lake trout, and brook trout) and two C. auratus (silver-E and color-A1) sequences, a linearized tree was constructed (data not shown). Similarly, diversity of transferrin molecule in C. auratus was estimated to begin 15.5 ± 2.6 MYA by referring to the divergence of salmonid genera.
Discussion
Amino Acid Sequence Variation and Molecular Basis of C. auratus Transferrin Polymorphism
Molecular mechanism of transferrin variations has been studied in many species from fishes to mammals (Hershberger 1970; Welch and Langmead 1990; Carpenter and Broad 1993; Laurent and Rodellar 2001). The genetic basis of transferrin polymorphism has been preliminarily discussed in C. a. auratus by using a simple model of color crucian carp (Yang, Zhou, and Gui 2004). A hypothesis was put forward that mobility of transferrin variants was negatively relevant to their Mw on SDS-PAGE gel and negatively relevant to their pI on native PAGE gel.
In this study, more cDNA sequences of transferrin alleles have been cloned from the silver crucian carp and white crucian carp (table 2). The deduced transferrin variants showed variable aa sequences and distinct computed physicochemical properties (fig. 2; table 3). By using the previous hypothesis on polymorphism of transferrin in color crucian carp (Yang, Zhou, and Gui 2004), a schematic graphic has been given to interpret the more complicated transferrin phenotypes of the silver crucian carp (fig. 8). With respect to the transferrin phenotypes on 8% PAGE gel, the distinct transferrin bands have been explained as encoded by three transferrin alleles of Tfa, Tfb, and Tfc (Yang et al. 2001). However, five silver crucian carp transferrin variants have been identified in this study by their variable cDNA sequences (table 2). As to their theoretical Mw, the condition is: silver-D > (silver-A = silver-B = silver-E) > silver-C (table 3). Hypothetically, three bands of different mobility were to emerge on SDS-PAGE gel. However, only two bands could be distinguished for these silver crucian carp clones on 12% SDS-PAGE gel (Yang et al. 2001). Most probably silver-D was not distinguished from silver-A, -B, and -E for the limited discrimination of PAGE, and silver-C encoded the faster band shared by clone D and clone A (fig. 8A). Meanwhile, five deduced transferrin variants show three pI values in silver crucian carp; they were aligned as: (silver-D = silver-E) < (silver-A = silver-B) < silver-C (table 3). Three transferrin bands were thus predicted for silver crucian carp on 8% PAGE, which corresponded to, respectively, the three transferrin alleles assumed previously (Yang et al. 2001). Namely, band a had the largest anodal migration and was encoded by silver-D or silver-E, band b ran at the intermediate speed and was encoded by silver-A/B, and the slowest band, c, was the product of allele silver-C (fig. 8B).
FIG. 8. The schematic transferrin phenotypes of silver crucian carp under the hypothesis that mobility of transferrin variants was determined by their Mw in SDS-PAGE (A) or by their pI in native PAGE (B). Transferrin bands are shown by continuous lines except for the discontinuous silver-A in clone A of silver crucian carp for its non-detection in PAGE. The transferrin genotypes of silver crucian carp clones (table 2) are expressed such that the transferrin alleles are listed in order of larger electrophoretic migration. The lowercase letters a, b, and c show the transferrin alleles previously assumed (Yang et al., 2001)
Transferrin allele silver-A was detected from all the three silver crucian carp clones by PCR (table 2). Its encoded band b was observed for clones D and F of silver crucian carp on 8% PAGE gel; however, it was not detected for clone A (Yang et al. 2001; indicated by discontinuous lines in fig. 8B). Plausibly deleterious variation happened in gene of silver-A and the abnormal expression resulted in its non-detection on PAGE gel, considering that the silver crucian carp was a unisexual and polyploid animal and tended to accumulate more mutations in its genome (Muller 1964; Risinger and Larhammar 1993).
Transferrin polymorphism was also observed between the two white crucian carp specimens (data not shown). Considerable variations were observed among the three transferrin alleles identified from the white crucian carp (fig. 2), but their genetic relationships with the transferrin variants have not been determined. Particularly, one potential N-glycosylation site was identified for transferrin sequences of white-A and white-B (fig. 2). Consequently, a more complicated condition has to be considered in the further discussion of genetic basis of transferrin polymorphism in the C. auratus.
Identification of Selected Transferrin Sites and Their Function Significance
Codons subject to positive selection had been identified in salmonid transferrins, and their homologous sites displayed the directional distribution in the peptides for binding of bacterial proteins in the bovine lactoferrin (Ford 2001). Polymorphic residues also showed a tendency to locate on external parts of transferrin variants in horse (Carpenter and Broad 1993). However, testing of 2 distribution refuted any association of the sites homologous to those selected in C. auratus transferrin with the peptides for bacterial binding in rabbit transferrin. In other words, the rapid evolution of transferrin alleles of C. auratus did not seem to be driven by the competition of iron by the pathogenic proteins.
In contrast to their irregular distribution in the primary structure of transferrin, however, five of the 11 homologous selected residues converged in a plane closely parallel to the cleft between the N1 and N2 domains (fig. 8). Studies of the crystal structure of human lactoferrin showed the probable condition that iron and anion were bonded to transferrin within the interdomain cleft (Anderson et al. 1989). The iron atom is coordinated to four protein ligands and the specific CO32- bound to iron as a bidentate bond. A few polar residues around the iron-binding cavity form hydrogen bonds with the anion or the iron, and more ones provided a hydrophilic environment for transferrin to bond to iron or anion. The five selected sites converged in the interdomain of the N-lobe of C. auratus transferrin could be replaced as the aa of Glu, Asp, Arg, His, and Cys (see fig. 2 for the possible replacements). These polar residues had potentiality to form hydrogen bonds with those directly bonded to iron and anion, and their situation in providing a hydrophilic cavity could be expected. Consequently, variation of these five selection sites might indirectly influence the capability of C. auratus transferrin to bind to or release iron. The other five correspondent selected sites appeared on the surface of rabbit transferrin and were not even near the interdomain cleft for binding of ions. However, these five residues were all located at linkages of different secondary structures (fig. 9); variations of them might result in the opening or closing of interdomain clefts or half molecules to different extents and ultimately might have an effect on the binding or releasing of iron (Anderson et al. 1989).
FIG. 9. Secondary structure of the diferric rabbit serum transferrin molecule (Hall et al. 2002). Helixes and strands are respectively indicated by cylinders and arrows, respectively. Residues with the probable function with iron or anion and those homologous to the selected sites in transferrin of C. auratus are indicated by white balls and black balls, respectively
These analyses implied that evolution of the selected sites in transferrin of C. auratus could probably interfere with its function in the uptake of iron. Iron is an essential element for the body because it is necessary for normal conformation and function of some iron-carrying proteins, such as hemoglobin and myoglobin. As a result, metabolisms of iron and oxygen are associated closely in the organism. For instance, expression of transferrin and its acceptor were simultaneously activated in the organism in the condition of hypoxia (Rolfs et al. 1997; Tacchini et al. 1999). The genus Carassius has long been known for its capability being of anoxia-tolerant (Lutz, Nilsson, and Perez-Pinzon 1996). Furthermore, C. auratus inhabits a vast territory of the Eurasia continent and has to experience rough environments that include drought or cold, both coupled with hypoxia. Consequently, diversification of C. auratus transferrin was plausibly firmly driven by these extreme environments, and the resulted abundant transferrin genotypes might provide better survival of C. auratus. Presumably, research on anoxia-tolerance of C. auratus individuals with distinct transferrin phenotypes might provide further proof of this hypothesis.
Phylogenetic Relationship and Antique Origin of C. auratus Transferrin Alleles
Two phylogenetic trees were constructed based on different regions of coding sequences of C. auratus transferrin (fig. 6; data not shown). Both phylogenies inferred the closer phylogenetic relationship between transferrins of silver crucian carp and color crucian carp, whereas the transferrin alleles of white crucian carp tended to form a monophylogeny. Remarkably, the highest sequence homology was found between the transferrin alleles of silver-C and color-B1, which were from two C. auratus subspecies with differentiation in both morph and habitat (Luo and Yue 2000). Besides these phylogenetic analyses, recombination events were occasionally observed between transferrin alleles of silver crucian carp and color crucian carp (fig. 5C). These data inferred the significant influence of geographic distribution on differentiation of three C. auratus populations. In other words, the white crucian carp was subject to more complete genetic isolation from other C. auratus populations on the separate Japanese islands. By contrast, silver crucian carp and color crucian carp inhabited the Asian continent and partially shared their gene pool. Hybridization introgression could occasionally happen between these two populations with recombination observed.
Molecular clock calculation inferred an evolutionary history of up to 15–20 Myr for the C. auratus transferrin alleles. The tetraploidization event in cyprinids was deduced to occur 16–22 MYA, estimated by divergence of the synaptosome-associated protein loci (Risinger and Larhammar 1993), and 50–70 MYA based on hemoglobin data (Powers and Edmundson 1972). Although it was unclear if these transferrin gene lineages emerged as the direct result of the polyploidization event, it might exert significant influence on their further diversification (see the next section). Some data (Yang, Zhou, and Gui 2004) has supported the allelic origin of the transferrin copies. They are considered as alleles in the following sections. However, further studies on their genomic localization and gene structure are necessary to clarify their origin.
Evolutionary Mechanism of High Transferrin Polymorphism in C. auratus
DNA polymorphism of transferrin has been studied in mammals (Carpenter and Broad 1993; Laurent and Rodellar 2001) and fishes (Ford 2000; Antunes et al. 2002). About 20 nucleotide sites were identified as variable among transferrin alleles, and fewer aa replacements were detected within these species. However, up to 248 variable nucleotide sites have been identified among the 10 transferrin alleles of C. auratus in this study. The average sequence identity of 89.5% among transferrin variants was even comparable to that among salmonid species (Lee et al. 1998). Then how has this redundant DNA polymorphism of transferrin happened?
The ancient allelic lineages of C. auratus might give the first interpretation. It was known that speciation of the C. auratus was not too long before 500 thousand years ago, when progenitor white crucian carp emerged (Murakami, Matsuba, and Fujitani 2001). Consequently, DNA polymorphism of C. auratus transferrin most likely derived earlier than the species C. auratus itself and was a transspecies polymorphism (Klein 1986). Furthermore, generally these allelic lineages did not diverge earlier than the tetraploidization event of cyprinids. The gene duplication coupled by polyploidization might provide doubled or even tripled (for the silver crucian carp; Zhu and Jiang 1993) genetic units for further diversification, such as recombination and selection.
Detection of codons subject to positive selection presented the second explanation. Most of the duplicated genes were silenced within a few Myr, and even the few survivors would probably experience strong purifying selection (Lynch and Conery 2000). However, positive selection could still been observed for some genes and played the important role in diversification of them for its accelerative fixation of replacement mutations (Hughes and Nei 1988; Yang 1998). Natural selection exerted on partial codons of C. auratus transferrin could elevate their aa substitutions and cause significant effects on its allelic diversification.
Furthermore, recombination was thought to play a role in the generation of novel alleles at various human major histocompatibility complex loci (Carrington 1999). It probably played the similar role in the generation of genetic variations in C. auratus transferrin, because four of its alleles showed the apparent recombination origin (fig. 4). Remarkably, sex and recombination accelerate the accumulation of beneficial mutation and reduce accumulation of harmful mutation by increasing realistic selection strength from background trapping (Rice and Chippindale 2001). Therefore, recombination and selection might interact with each other in increasing genetic variation of C. auratus transferrin.
The antique allelic lineages of C. auratus transferrin strongly implied a balancing selection mechanism in the maintenance of their polymorphism, as neutral alleles could only persist in a population for limited generations after their fixation (Takahata and Nei 1990).
Two forms of balancing selection might then happen, overdominant or frequency-dependent, with the former preferable for heterozygotes and the latter tending to retain the genotype of lower allele frequencies in the population. High transferrin heterozygosity has been observed to be prevalent in different fishes (Kirpichnikov 1981). Zhang (1996) investigated the transferrin polymorphism in a wild C. auratus population (C. auratus auratus). Most of the individuals (24 vs. 26) showed the heterozygous transferrin phenotypes. Furthermore, 18 of the 28 individuals of the color crucian carp were transferrin heterozygotes in our studies (Yang, Zhou, and Gui 2004), as well as were all the silver crucian carp clones (Yang et al. 2001). Collectively, these data suggested a heterozygosity not less than 50% in the C. auratus population. Furthermore, a significant characteristic of overdominant selection might be the detection of positive selection for its existence as a strong drive for novel alleles with substantial variations and higher heterzygosity in the population (Li 1997, pp. 237–268). Positive selection was evidently confirmed for transferrin alleles of C. auratus, and the sites subject to selection were also identified. Based on these results, overdominance might be the more possible selective mechanism for long persistence of transferrin polymorphism in the population. However, direct data of comparison of fitness among transferrin phenotypes and investigation of allele frequency distribution were requisite to test this type of balancing selection.
The silver crucian carp is a special C. auratus population with complicated evolutionary backgrounds, which might exert significant influence on the selective mechanism of its transferrin. With the clonal reproductive mode without recombination, mutation tends to rapidly fix in the clone. Most of the mutations are neutral in expectation. However, the silver crucian carp clone with a new mutation is most probably a heterozygote for the asymmetry of mutation, which has the higher fitness under both types of balancing selection. Subsequently, this clone will propagate quickly without the "cost of males" (Smith 1975) and will completely occupy its niche. Once the environment has changed, fitness of this clone will lower to moderate or poor. The old transferrin allele will be lost easily in genetic drift or even purged by negative selection because it cannot participate in the creation of a novel genotype for absence of recombination. Apparently, positive selection can be easily observed in the clonal population of silver crucian carp, but the long persistence of transferrin polymorphism seems impossible.
Interestingly, silver crucian carp is never an asexual population absolutely separate in the nature. A minority ratio of males was found in the silver crucian carp populations and often lived with its bisexual related populations, such as C. a. auratus, Carassius carassius, and common carp (Cherfas 1981; Fan and Shen 1990). Significantly, sexual reproduction and recombination have been observed among silver crucian carp clones (Zhou, Wang, and Gui 2000; Zhou and Gui 2001). In the current study, we have also revealed the recombination events among their transferrin alleles (fig. 5).
Vrijenhoek's frozen niche variation model (1979) could be used to explore the realistic balancing selection mechanism in the transferrin of silver crucian carp. As is the condition in other unisexual vertebrates, the silver crucian carp probably has a hybridization origin (Cherfas 1981; Yang et al. 2001). The silver crucian carp clones continually capture and freeze genetic variation from their sexual ancestors, and multiple heterozygotic transferrin genotypes are therefore fixed. Selection among clones favors specialized genotypes having minimal niche overlap with other clones and the sexual ancestors (the marginal overdominance; Wallace 1968). Furthermore, the novel transferrin alleles and genotypes can come from hybridization between silver crucian carp clones or silver crucian carp and their related bisexual populations, which are fixed again in the form of clone (Zhou, Wang, and Gui 2000).
Presumably, the versatile balancing selection type enables silver crucian carp to survive in complicated environments. Stronger positive selection can be expected in the silver crucian carp clones for the rapid fixation of advantageous mutations in gynogenetic clones of silver crucian carp (fig. 3, the silver-A/silver-B cluster). Recombination can not only create abundant genetic variations, but it can also remove the accumulating mutational load in the asexual population (Rice 1998). Significantly, novel alleles can be introduced into the silver crucian carp clones for their hybridization origin, and the obsolete alleles may participate in the creation of new transferrin phenotypes by occasional sexual reproduction, which enables the persistence of ancient allelic lineages in the population. Therefore, owing to its particular evolutionary backgrounds, the silver crucian carp might possess a distinctive balancing selection mechanism.
FIG. 6. Neighbor-joining tree for transferrins of chicken, salmonids and C. auratus, with human assumed as the outgroup. The branch lengths (see scale at bottom) are measured in terms of the number of nucleotide substitutions per site based on the middle-region coding sequences of transferrins (332 nucleotides). Numbers given to interior branches are the bootstrap values higher than 50% with 1,000 replicate resamplings (Kumar et al. 2001). Three salmonid genera, Salmo, Salvelinus, and Oncorhynchus, are indicated by brackets. Asterisks (**) indicate the transferrin alleles evolving significantly faster at the 1% level in the branch-length test (Takezaki, Rzhetsky, and Nei 1995)
Acknowledgements
This work was supported by the National Natural Science Foundation of China (grant numbers 30200028, 30130240, and 30123004), the Chinese Academy of Sciences (KSCX2-SW-303, KSCX2-1–05), the Institute of Hydrobiology of the Chinese Academy of Sciences (220309), and the European Commission (contract no. ICA4-CT-2001–10024).
Literature Cited
Anderson, B. F., H. M. Baker, G. E. Norris, D. W. Rice, and E. N. Baker. 1989. Structure of human lactoferrin: crystallographic structure analysis and refinement at 2.8 ? resolution. J. Mol. Biol. 209:711-734.
Antunes, A., A. R. Templeton, R. Guyomard, and P. Alexandrino. 2002. The role of nuclear genes in intraspecific evolutionary inference: genealogy of the transferring gene in the brown trout. Mol. Biol. Evol. 19:1272-1287.
Arbogast, B. S., S. V. Edwards, J. Wakeley, P. Beerli, and J. B. Slowinski1. 2002. Estimating divergence times from molecular data on phylogenetic and population genetic time scales. Annu. Rev. Ecol. Syst. 33:707-740.
Baldwin, G. S. 1993. Comparison of transferrin sequences from different species. Comp. Biochem. Physiol. 106B:203-218.
Carpenter, M. A., and T. E. Broad. 1993. Polymorphism in the coding sequence of the horse transferrin gene. Genome 37:157-164.
Carrington, M. 1999. Recombination within the human MHC. Immunol. Rev. 167:245-256.
Cherfas, N. B. 1981. Gynogenesis in fishes. Pp. 255–273 in V. S. Kirpichnikov, ed. Genetic bases of fish selection. Springer-Verlag, Berlin.
Fan, Z., and J. Shen. 1990. Studies on the evolution of bisexual reproduction in crucian carp (Carassius auratus gibelio Bloch). Aquaculture 84:235-244.
Felsenstein, J. 1993. PHYLIP (phylogeny inference package). Version 3.5c. Distributed by the author, Department of Genetics, University of Washington, Seattle.
Ford, M. J. 2000. Effects of natural selection on patterns of DNA sequence variation at the transferrin, somatolactin, and p53 genes within and among chinook salmon (Oncorhynchus tshawytscha) populations. Mol. Ecol. 9:843-855.
Ford, M. J. 2001. Molecular evolution of transferrin: evidence for positive selection in salmonids. Mol. Biol. Evol. 18:639-647.
Frelinger, J. A. 1972. The maintenance of transferrin polymorphism in pigeons. Proc. Natl. Acad. Sci. USA 69:326-329.
Gui, J. F. 1996. A unique study system: gynogenetic fish Carassius auratus gibelio. Science Foundation of China 4:44-46.
Gui, J. F., S.C. Liang, L. F. Zhu, and Y. G. Jiang. 1993. Discovery of two different reproductive development modes of the eggs of artificial multiple tetraploid allogynogenetic silver crucian carp. Chinese Sci. Bull. 38:332-337.
Hall, D. R., J. M. Hadden, G. A. Leonard, S. Bailey, M. Neu, M. Winn, and P. F. Lindley. 2002. The crystal and molecular structures of diferric porcine and rabbit serum transferrins at resolutions of 2.15 and 2.60 ?, respectively. Acta Crystallogr. D Biol. Crystallogr. 58:70-80.
Hershberger, C. L., J. L. Larson, and B. Arnold, et al. (12 co-authors). 1991. A cloned gene for human transferrin. Ann. N.Y. Acad. Sci. 646:140-154.
Hershberger, W. K. 1970. Some physicochemical properties of transferrins in brook trout. Trans. Amer. Fish. Soc. 1970:207-218.
Hey, J., and J. Wakeley. 1997. A coalescent estimator of the population recombination rate. Genetics 145:833-846.
Hudson, R. R., and N. L. Kaplan. 1985. Statistical properties of the number of recombination events in the history of a sample of DNA sequences. Genetics 111:147-164.
Hughes, A. L., and M. Nei. 1988. Pattern of nucleotide substitution at major histocompatibility complex class I loci reveals overdominant selection. Nature 335:167-170.
Jeltsch, J. M., and P. Chambon. 1982. The complete nucleotide sequence of the chicken ovotransferrin mRNA. Eur. J. Biochem. 122:291-295.
Kimura, M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. J. Mol. Evol. 16:111-120.
Kirpichnikov, V. S. 1981. The genetics of nonenzymatic proteins in fishes. Pp. 162–172 in V. S. Kirpichnikov, ed. Genetic bases of fish selection. Springer-Verlag, Berlin.
Klein, J. 1986. Natural history of the major histocompatibility complex. John Wiley, New York.
Kumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001. MEGA2: molecular evolutionary genetics analysis software. Bioinformatics 17:1244-1245.
Kvingedal, A. M., K. A. Rorvik, and P. Alestrom. 1993. Cloning and characterization of Atlantic salmon (Salmo salar) serum transferrin cDNA. Mol. Mar. Biol. Biotechnol. 2:233-238.
Laurent, P., and C. Rodellar. 2001. Characterization of a single nucleotide polymorphism in the coding sequence of the bovine transferrin gene. Mut. Res. Genomics 458:1-5.
Lee, J. Y., T. Tada, I. Hirono, and T. Aoki. 1998. Molecular cloning and evolution of transferrin cDNAs in salmonids. Mol. Mar. Biol. Biotech. 7:287-293.
Lee, J.-Y., N. Tange, H. Yamashita, I. Hirono, and T. Aoki. 1995. Cloning and characterization of transferrin cDNA from coho salmon (Oncorhynchus kisutch). Fish Pathol. 30:271-277.
Li, W. H. 1997. Molecular evolution. Sinauer Associates, Sunderland, Mass.
Luo, Y. L., and P. Q. Yue. 2000. Cyprininae. Pp. 427–433 in P. Q. Yue et al, eds. Fauna sinica, Osteichthyes, Cypriniformes III. Science Press, Beijing.
Lutz, P. L., G. E. Nilsson, and M. A. Perez-Pinzon. 1996. Anoxia tolerant animals from a neurobiological perspective. Comp Biochem Physiol 113B:3-13.
Lynch, M., and J. S. Conery. 2000. The evolutionary fate and consequences of duplicate genes. Science 290:1151-1155.
Muller, H. J. 1964. The relation of recombination to mutational advance. Mutat. Res. 1:2-9.
Murakami, M., C. Matsuba, and H. Fujitani. 2001. The maternal origins of the triploid ginbuna (Carassius auratus langsdorfi): phylogenetic relationships within the C. auratus taxa by partial mitochondrial D-loop sequencing. Genes Genet. Syst. 76:25-32.
Nielsen, R., and Z. Yang. 1998. Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene. Genetics 148:929-936.
Ohno, S. 1970. Evolution by gene duplication. Springer, Berlin.
Powers, D. A., and A. B. Edmundson. 1972. Multiple hemoglobins of catostomid fish. II. The amino acid sequence of the major chain from Catostomus clarkii hemoglobins. J. Biol. Chem. 247:6694-6707.
Retzer, M. D., R. H. Yu, and A. B. Schryvers. 1999. Identification of sequences in human transferrin that bind to the bacterial receptor protein, transferrin-binding protein B. Mol. Microbiol. 32:111-121.
Rice, W. R. 1998. Requisite mutational load, pathway epistasis, and deterministic mutation accumulation in sexual versus asexual populations. Genetica 102/103:71-81.
Rice, W. R., and A. K. Chippindale. 2001. Sexual recombination and the power of natural selection. Science 294:555-559.
Risinger, C., and D. Larhammer. 1993. Multiple loci for synapse protein SNAP-25 in the tetraploid goldfish. Proc. Natl. Acad. Sci. USA 90:10598-10602.
Robertson, D. L., B. H. Hahn, and P. M. Sharp. 1995. Recombination in AIDS viruses. J. Mol. Evol. 40:249-259.
Rolfs, A., I. Kvietikova, M. Gassmann, and R. H. Wenger. 1997. Oxygen-regulated transferrin expression is mediated by hypoxia-inducible factor-1. J. Biol. Chem. 272:20055-20062.
Smith, J. M. 1975. Theory of evolution. 3rd edition. Harmondsworth: Penguin.
Tacchini, L., L. Bianchi, A. Bernelli-Zazzera, and G. Cairo. 1999. Transferrin receptor induction by hypoxia: HIF-1-mediated transcriptional activation and cell-specific post-transcriptional regulation. J. Biol. Chem. 274:24142-24146.
Takahata, N., and M. Nei. 1990. Allelic genealogy under overdominant and frequency-dependent selection and polymorphism of major histocompatibility complex loci. Genetics 124:967-978.
Takezaki, N., A. Rzhetsky, and M. Nei. 1995. Phylogenetic test of the molecular clock and linearized trees. Mol. Biol. Evol. 12:823-833.
Tange, N., J.-Y. Lee, N. Mikawa, I. Hirono, and T. Aoki. 1997. Cloning and characterization of transferrin cDNA and rapid detection of transferrin gene polymorphism in rainbow trout (Oncorhynchus mykiss). Mol. Mar. Biol. Biotechnol. 6:354-359.
Thompson, J. D., T. J. Gibson, F. Plewniak, F. Jeanmoughin, and D. G. Higgins. 1997. The ClustalX windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 24:4876-4882.
Vrijenhoek, R. C. 1979. Factors affecting clonal diversity and coexistence. Am. Zool. 19:787-797.
Wallace, B. 1968. Topic in population genetics. W. W. Norton, Inc., New York.
Wei, W. H., J. Zhang, Y. B. Zhang, L. Zhou, and J. F. Gui. 2003. Genetic heterogeneity and ploidy level analysis among different gynogenetic clones of polyploid gibel carp. Cytometry 56:46-52.
Welch, S., and L. Langmead. 1990. A comparison of the structure and properties of normal human transferrin and a genetic variant of human transferrin. Int. J. Biochem. 22:275-282.
Williams, J. 1968. A comparison of glycopeptides from the ovotransferrin and serum transferrin of the hen. Biochem. J. 108:57-67.
Wilson, M. R., D. Middleton, and G. W. Warr. 1988. Immunoglobulin heavy chain variable region gene evolution: structure and family relationships of two genes and a pseudogene in a teleost fish. Proc. Natl. Acad. Sci. USA 85:1566-1570.
Yang, L., S. T. Yang, X. H. Wei, and J. F. Gui. 2001. Genetic diversity among different clones of the gynogenetic silver crucian carp, Carassius auratus gibelio, revealed by transferrin and isozyme markers. Biochem. Genet. 39:213-225.
Yang, L., L. Zhou, and J. F. Gui. 2004. Molecular basis of transferrin polymorphism in goldfish (Carassius auratus). Genetica (in press).
Yang, Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. CABIOS 13:555-556.
Yang, Z. 1998. Likelihood ratio tests for detecting positive selection and application to primate lysozyme evolution. Mol. Biol. Evol. 15:568-573.
Yi, M. S., Y. Q. Li, J. D. Liu, L. Zhou, Q. X. Yu, and J. F. Gui. 2003. Molecular cytogenetic detection of paternal chromosome fragments in allogynogenetic gibel carp, Carassius auratus gibelio. Bloch. Chromosome Res. 11:665-671.
Yu, X. J., T. Zhou, Y. C. Li, K. Li, and M. Zhou. 1989. Chromosomes of Chinese fresh-water fishes. Science Press, Beijing.
Zhang, H. 1996. Studies on the genetic population structure of aquatic animals. Doctoral dissertation, Chinese Academy of Sciences.
Zhou, L., and J. F. Gui. 2001. Studies on genetic diversity in gonochoristic offspring produced from mating between two different gonogenetic clones of silver crucian carp. Acta Biologiae Experimentalis Sinica 34:169-176.
Zhou, L., and J. F. Gui. 2002. Karyotypic diversity in polyploid gibel carp, Carassius auratus gibelio Bloch. Genetica 115:223-232.
Zhou, L., Y. Wang, and J. F. Gui. 2000. Genetic evidence for gonochoristic reproduction in gynogenetic silver crucian carp (Carassius auratus gibelio Bloch) as revealed by RAPD assays. J. Mol. Evol. 51:498-506.
Zhu, L. F., and Y. G. Jiang. 1993. A comparative study of the biological characters of gynogenetic clones of silver crucian carp (Carassius auratus gibelio). Acta Hydrobiologica sinica 17:112-120.(Lin Yang*, and Jian-Fang )
Modern Virology Research Center, College of Life Sciences, Wuhan University, Wuhan, China
E-mail: jfgui@ihb.ac.cn.
Abstract
Transferrin polymorphism has been studied in the polyploid Carassius auratus by cloning and sequence analysis of cDNAs from its three subspecies C. auratus gibelio, C. auratus auratus, and C. auratus cuvieri. DNA polymorphism of extremely high extent was shown for the transferrin gene by the 248 segregation sites among coding region sequences of its alleles. The deduced amino acid sequences of the transferrin alleles showed variable theoretical physicochemical parameters, which might constitute molecular basis for their electrophoretic heterogeneity. Positive selection was inferred by the replacement/synonymous ratios larger than 1 in partial allelic lineages which was subsequently confirmed by likelihood simulation under neutral or selection models. Furthermore, the correspondent sites to these selected codons were collectively located at two planes in the crystallographic structure of rabbit transferrin, which suggested that the rapid evolution of C. auratus transferrin might correlate to its adaptation to variable environmental elements such as oxygen pressure. The minimal 26 recombination events were detected among coding sequences of C. auratus transferrin, with partial mosaic sequences and breakpoints identified by identity scanning and information site analyses. Phylogenetic analyses revealed multiple antique allelic lineages of transferrin, which was estimated to diverge fifteen to twenty MYA. All these features strongly suggested the role of balancing selection in long persistence of high transferrin polymorphism in C. auratus. Furthermore, owing to its particular evolutionary backgrounds, the silver crucian carp might possess a distinctive balancing selection mechanism.
Key Words: transferrin ? polymorphism ? positive selection ? balancing selection ? allele ? recombination ? Carassius auratus
Introduction
Transferrin is a protein involved in iron storage and transport of vertebrates. Abundant polymorphism was observed in electrophoretic phenotypes of transferrin in many species, and its zymograms were often used as genetic markers for studying of population genetics (Kirpichnikov 1981). Meanwhile, many efforts have been made to elucidate the genetic basis of this polymorphism. In human, transferrin variants of different electrophoretic migrations have been successively identified, and most of them exhibit similar physiochemical properties and functions despite the occasional replacements among their amino acid (aa) sequences (Welch and Langmead 1990). In horse and cattle, transferrin polymorphism was studied at the gene level, and a series of single nucleotide polymorphism sites and aa substitutions respectively were identified. This had partially interpreted the heterogeneous phenotypes of different transferrin variants (Carpenter and Broad 1993; Laurent and Rodellar 2001). Furthermore, posttranslational glycosylation was also characterized in some species responsible for electrophoretic variation of transferrin (Williams 1968).
In studying transferrin polymorphism, attempts have been made to elucidate its evolutionary mechanism. The often high level of polymorphism in transferrin was thought to be maintained by a preferable selection for heterozygotes based on the observation of heterozygous advantage or high heterozygosity within populations (Frelinger 1972; Kirpichnikov 1981). Ford (2000, 2001) studied molecular evolution of transferrin in the salmonids. Positive selection was confirmed to play significant roles in diversification of transferrin molecules among salmonid species, and localization of the selected sites suggested the competition for iron from pathogenic bacterial as the evolutionary drive (Ford 2001). Recently, evidences for positive selection of tranferrin were further revealed at the intraspecific level in the brown trout, and recombination and gene conversion events responsible for diversification of transferrin alleles were also detected (Autunes et al. 2002).
In cyprinid fish, the species Carassius auratus and its subspecies have been widely recognized as model animals in evolutionary biology (Ohno 1970; Gui 1996; Zhou and Gui 2002). The common goldfish (C. auratus auratus) has a distribution throughout the Eurasia continent. It has 100± chromosomes (Yu et al. 1989) and multiple duplicated loci were identified from its genome (Wilson, Middleton, and Warr 1988; Risinger and Larhammar 1993), which inferred a probable tetraploidization event in its earlier evolution. The white crucian carp (C. auratus cuvieri), originating from Lake Biwa of Japan, also has 100± chromosomes (Gui et al. 1992). The silver crucian carp (C. auratus gibelio) has a distribution over a large territory over Europe, Russia, China and Japan. It has over 150 chromosomes and higher ploidy (Zhu and Jiang 1993; Zhou and Gui 2002; Wei et al. 2003; Yi et al. 2003). Two different reproductive modes were revealed in this subspecies, i.e., allogynogenetic reproduction and gonochoristic reproduction (Gui et al. 1993; Zhou, Wang, and Gui 2000; Zhou and Gui 2001). Transferrin polymorphism has been observed in the silver crucian carp, and three distinct transferrin phenotypes have been revealed among its three gynogenetic clones: A, D, and F (Yang et al. 2001). Three alleles were considered to genetically determine this polymorphism, and the heterozygotic transferrin genotypes apparently showed a clonal inheritance pattern. In a color variety of C.a. auratus, color crucian carp, the three transferrin variants A1, A2, and B1 have been identified, and substantial substitutions were observed among their alleles (Yang, Zhou, and Gui 2004).
In this study, we further cloned eight alleles of transferrin from the silver crucian carp and white crucian carp. Sequence variations among transferrin alleles of the polyploid C. auratus were studied comparatively, and positive selection and recombination were verified to play important roles in their diversification. Furthermore, multiple antique allelic lineages were identified, which inferred a balancing selection mechanism in maintenance of their polymorphism.
Materials and Methods
Specimens Collection and Electrophoretic Analysis of Transferrin Phenotypes
Three silver crucian carp individuals were selected for experiments, respectively representing the three silver crucian carp gynogenetic clones of A, D, and F (Yang et al. 2001). Two white crucian carp specimens were collected from an artificially maintained small population from the Duofu Farm. The common carp individual was identified to be from the Xingguo red common carp strain (Cyprinus carpio red variety). All of these six fish specimens had been raised in separate ponds of the Guanqiao Experimental Station, the Institute of Hydrobiology, the Chinese Academy of Sciences. The three silver crucian carp clones were identified by polyacrylamide gel electrophoresis (PAGE) analysis of their sera (Yang et al. 2001). Serum transferrin of the two white crucian carp individuals was also analyzed, and they showed different phenotypes.
Amplification of Transferrin cDNA by Polymerase Chain Reaction (PCR)
Total RNA was isolated from the fresh livers of the six fish specimens using Trizol reagent (GIBCO/BRL), and cloning of transferrin cDNA had begun from clone D of gynogenetic silver crucian carp. Degenerated primers Rt-5'/Rt-3', selected from the coding region of the transferrin gene in other vertebrates, were used for amplifying an original transferrin cDNA fragment (table 1). First strand cDNA was reverse-transcribed from 2 μg liver RNA in a 20-μl reaction mixture containing 200 u of M-MuLV reverse transcriptase (GIBCO BRL). One or two μl of this cDNA was used for amplification in a 50-μl reaction mixture containing 1 x buffer, 200 μM dNTP, 0.5 μM of each primer, and 2 units of Taq DNA polymerase (Biostar). The cycling profile was 1 cycle of 95°C for 5 min, 54°C for 2 min, and 72°C for 40 min followed by 30 cycles of 94°C for 1 min, 48°C for 1min, and 72°C for 3 min. Cloning and sequencing analysis confirmed this amplified product as the transferrin cDNA fragment. Furthermore, detection of two clones with the variable nucleotide sequences (designated as silver-A and silver-B) suggested the probable existence of multiple transferrin alleles (fig. 1). Gsp-5' and Gsp-3' primers were then selected from these sequences for amplification of the cDNA ends. First strand cDNA was initiated by primer Seq-3', recovered by use of NucleoTrap PCR Purification Kit (CLONTECH), and then added with a poly(A)+ tail using terminal transferase (Promega). Four μl of this cDNA was used for amplification of the 5' transferrin cDNA ends in a 50-μl cocktail containing 0.1 μM Adap-(T)17 and 0.5 μM Gsp-5'. 0.5 μM of Adap primer was added after the initial cycle of 95°C for 5 min, 43°C for 2 min, and 72°C for 40 min. This was followed by 30 cycles of 94°C for 1 min, 58°C for 1 min, and 72°C for 1 min. Amplification of the 3' cDNA ends was performed as described for Rt-5'/Rt-3' amplification except that Adap was used for the downstream PCR primer.
Table 1 Primers Used for the Amplification of C. auratus Transferrin cDNA.
FIG. 1. Amplification of transferrin cDNA sequences from D clone of gynogenetic silver crucian carp. (top section) Primary structure of crucian carp transferrin cDNAs, with shaded box and lines displaying the open reading frame and the untranslated regions, respectively. (bottom section) Three approximately complete transferrin cDNA sequences were amplified. First, a transferrin cDNA fragment was amplified by PCR with degenerated primers Rt-5'/Rt-3', with alleles silver-A and silver-B identified. Then, 5' and 3' cDNA ends were PCR amplified, but a complete cDNA sequence could not be presented for the variable Rt-5'/Rt-3' sequences. Thus, a nearly complete transferrin cDNA sequence had to be amplified by gene-specific For/Rev, by which the third transferrin allele, silver-C, was identified. Allele-specific PCR were then employed to amplify cDNA ends of silver-A or silver-B, and their complete cDNA sequences were thus obtained. The pattern of strand indicates cDNA sequences of different transferrin alleles: black, of silver-A; half-transparent, of silver-B; and transparent, of silver-C. The numbers of sequenced clones are listed on the PCR products. An asterisk (*) indicates that the cDNA fragment was from the silver-A' transferrin allele, with sequence identical to that of silver-A in this region
Unexpectedly, no sequence variation was detected among clones from the 5' or 3' transferrin cDNA ends; thus, a complete transferrin cDNA sequence could not be presented by splicing of the three available PCR products (fig. 1). The gene-specific primer set For/Rev was selected from the cDNA ends to amplify a complete transferrin cDNA sequence directly, by which the third transferrin allele, silver-C, was identified. Afterwards, primers specifically annealing to cDNA sequences of certain alleles such as Var-5'(A) and Var-3'(A) were employed to amplify their 5' and 3' ends (table 1); ultimately, their integrate cDNA sequences were obtained.
With similar procedures, variable transferrin cDNA sequences were amplified from liver RNA of silver crucian carp clone A and clone F; two other alleles, silver-D and silver-E, were then identified from these two carp clones, respectively. With respect to white crucian carp, cloning of transferrin cDNA sequence was accomplished by sequencing of four clones of For/Rev PCR products, with two from the two individuals each, by which three transferrin alleles, white-A, white-B and white-C, were detected. Detection of transferrin alleles from the six C. auratus individuals is listed in table 2.
Table 2 Detection of Transferrin Alleles Among C. auratus Individuals by RNA PCR.
Cloning and Sequence Determination of PCR Products
PCR products were isolated by agarose gel electrophoresis using NucleoTrap Gel Extraction Kit (CLONTECH). Recovered DNA fragments were directly cloned into pGEM-T vector as the manufacturer described (Promega). One to three clones were subjected to sequence determination for each of the PCR products, except that eight ones were sequenced for Rt-5'/Rt-3' product to detect more plausible transferrin alleles (refer to fig. 1 for the condition in clone D). Sequence analysis was programmed using dRhodamine terminator cycle sequencing Kit on ABI PRISMTM 310 Genetic Analyzer (Perkin Elmer). Translation of aa sequences, predication of signal peptide, and computation of isoelectric point and molecular weight (pI/Mw) were accomplished by use of software at ExPASy Molecular Biology Server (http://expasy.pku.edu.cn).
Estimation of dn/ds Ratios and Selected Sites
Besides the eight transferrin coding sequences presented here (table 3), two cDNA sequences had been described for color crucian carp previously (GenBank accession numbers AF518746 and AF518747; Yang, Zhou, and Gui 2004). All these sequences were aligned using the ClustalX 1.8 program (Thompson et al. 1997) with default alignment parameters (gap opening penalty 10.00; gap extension penalty 0.20; sequences >30% diverged delayed; DNA transition weight penalty 0.50; with negative matrix off). The obtained sequence alignments were 1,965 base pairs long, without complete 5' ends and any gaps.
Table 3 Theoretical Physiochemical Properties of Transferrin Variants from C. auratus.
For analysis of the plausible selective drives in the evolution of transferrin, a phylogenetic tree was constructed from the aligned DNA sequences using the maximum-likelihood method in the DNAML program of PHYLIP3.5 package (Felsenstein 1993; see fig. 3 for detailed parameters). Furthermore, the PAML computer package (Yang 1997) was used to examine the possible evolutionary models of the transferrin of C. auratus. Four codon-based models—single-ratio, free-ratio, neutral, and selection—were successively applied on different molecular data (Nielsen and Yang 1998; Yang 1998). Of the many parameters involved, "model for codons" and "NSsites" were especially considered for their crucial effects on selection of evolutionary models. Following were their settings for the four given models in order: 0, 0; 1, 0; 0, 1; 0, 2. The fitness of different models was examined by considering the 2 distribution of 2l, twice the log likelihood difference between any of the two models, with the model of likelihood value significantly larger than the preferable one. In the selection model, codons subject to selection were identified by their high posterior probability in the empirical Bayes's approach (Nielsen and Yang 1998). Furthermore, the sites had to show a replacement substitution rate (dn) larger than the silent substitution rate (ds) for confirmation of their positive selection. The software Cn3d3.0 was available from NCBI (www.ncbi.nlm.nih.gov) for the illustration and disposal of the three-dimensional structure of rabbit transferrin.
FIG. 3. Maximum-likelihood phylogeny based on transferrin coding sequences of C. auratus. The tree topology was estimated using the DNAML program in the PHYLIP package (Felsenstein 1993). Numbers next to the branches are the dn/ds ratios estimated under the free-ratio model of PAML package (Yang 1998)
Recombination Analysis
After removal of the 11 selected codons (fig. 2), population recombination rates and the minimal set of recombination intervals (Hudson and Kaplan 1985) were estimated for transferrin sequences C. auratus (with 1,932 nucleotides left) using the program SITES version 1.1 (Hey and Wakeley 1997). Because incomplete reproduction isolation and apparent migration happened in the divergence of the three C. auratus subspecies, transferrin sequences from all of them and uniquely from the gynogenetic silver crucian carp were subject to calculation respectively for comparison.
FIG. 2. Alignment of putative aa sequences of transferrin from C. auratus and common carp (GenBank accession no. AF457152; Yang, Zhou, and Gui 2004). Transferrin alleles of C. auratus and their encoded proteins are abbreviated: sil_A for silver-A and sil-A for silver-A, respectively, etc. Dots (.) represent aa residues identical to those of transferrin variant A from the silver crucian carp. Gaps (-) are introduced to optimize identity. Conserved regions involved in the iron ligand and anion binding of transferrin are shaded. Asterisks (*) represent the iron or anion ligands. The predicted signal peptide sequences of transferrins are shown by underlined letters. The signal "#" marks the potential residues for glycosylation in white-A and white-B. Triangles indicate the sites subject to positive selection. The program ClustalX 1.8 was used for sequence comparison and alignment (Thompson et al. 1997)
The recombination intervals and breakpoints were then determined for transferrin alleles by use of the software SIMPLOTS2.5 (Robertson et al. 1995). The plausible mosaic sequences were first identified from coding sequences of the transferrin alleles by querying each one to the others using the Plot and BootScan functions. Four sequences were sorted out to form a group including the mosaic allele, the two plausible parental alleles, and a background allele. Informative sites were identified for these sequences with each one supporting one possible phylogeny. Locations of the breakpoints were then determined by their maximizing the 2 value among the phylogeny clusters. P-values for the resultant divisions of sites were calculated using Fisher's exact test.
Evolutionary Rate Tests and Phylogenetic Analysis
Recombination and selection affect the molecular clock (Arbogast et al. 2002); thus, divergence time estimation requires consideration of the interference of recombination and selection. Becuase recombination was detected primarily through the intermediate region of transferrin coding sequence, the homogenous regions of the 5' and 3' ends were selected for the molecular clock test and phylogeny reconstruction using the LINTREE program (Takezaki, Rzhetsky, and Nei 1995). A number of transferrin sequences were cloned from salmonids (refer to Ford 2001 for index), and divergence of these species might provide an ideal calibration point for molecular clock calculation. Nine of their transferrin sequences are cited here (with their GenBank accession numbers listed in parentheses): Atlantic salmon (L20313; Kvingedal, Rorvik, and Alestrom 1993), brown trout (D89091; Lee et al. 1998), lake trout (D89090; Lee et al. 1998), brook trout (D89089; Lee et al. 1998), Japanese Char (D89088; Lee et al. 1998), coho salmon (D89084; Lee et al. 1995), sockeye salmon (D89095; Lee et al. 1998), rainbow trout (D89083; Tange et al. 1997), amago salmon (D89086; Lee et al. 1998). Coding sequences of human (S95936; Hershberger et al. 1991) and chicken (X02009; Jeltsch and Chambon 1982) transferrin genes were also used as an outgroup. These 11 transferrin sequences were aligned with those of C. auratus using the default parameters of the ClustalX 1.8 program (Thompson et al. 1997). In the transferrin sites of C. auratus and salmonids subject to positive selection (Ford 2001), the 5' obscure nucleotide and all the gaps were removed from the sequence alignments when a 332-nucleotide alignment and a 530-nucleotide alignment were obtained for the middle region and 3' end of transferrin sequences.
Neighbor-joining trees were constructed by LINTREE, with evolutionary distance measured by Kimura's two-parameter distance (Kimura 1980). Credibility of the nodes was assessed by the interior branch test from the MEGA2.0 program (Kumar et al. 2001). Two cluster tests examined the difference between distances within each interior node, and the branch length test examined the deviation of the branch length between the tree root and a tip from the average length (Takezaki, Rzhetsky, and Nei 1995). Both of these differences were denoted by , and its deviation from zero (for two-cluster test) or from the average branch length (for branch length test) was valued by the two-tailed normal deviate test. The heterogeneous sequences were eliminated from trees until all the nodes or branches showed the constant evolutionary rate (P > 1%). A linearized tree was then constructed by re-estimation of the branch lengths.
Results
Transferrin cDNAs and Their Deduced Amino Acid Sequences
Besides two transferrin cDNA sequences from the color crucian carp (Yang, Zhou, and Gui 2004), eight novel transferrin cDNAs have been cloned in this study. They include five alleles (silver-A, -B, -C, -D, -E) from three gynogenetic clones (D, A, F) of the silver crucian carp (table 2) and three alleles (white-A, -B, -C) from two individuals of white crucian carp.
The transferrin cDNAs showed sequences of 2,228 (white-B) to 2,372 (silver-A) nucleotides. Although only two of these sequences (silver-A and silver-C) had poly (A)+ tails in the 3' ends, all of them contained complete open reading frames as well as their putative aa sequences; they are available from GenBank (table 3).
The transferrin alleles also encode variable lengths of protein sequences: 661 aa for silver-C; 666 aa for silver-E; 669 aa for silver-A and -B; 670 aa for white-B; and 671 aa for silver-D and white-A and -C (fig. 2; table 3). Meanwhile, they contain two lengths of signal peptide sequences: 15 aa for silver-C, -D, and -E and white-A and -B; and 21 aa for silver-A and -B and white-C. In contrast to the studies in many other species (Baldwin 1993; Lee et al. 1998), most transferrin variants of C. auratus have no potential N-glycosylation site and do not seem to be glycoproteins (with a single modification site found for white-A and -B as the exception). Plausible variations could also be found in these variants, and 34 and 37 cysteine residues were mainly observed in silver crucian carp and white crucian carp transferrins, respectively. After removal of the presumed signal peptides, Mw and pI values were calculated. As shown in table 3, the predicted Mw vary in a limited range: silver-D and white-A have the largest one of about 72.5 kDa; white-B has the second largest of 72.2 kDa; silver-A, silver-B, silver-E, and white-C are 71.9 kDa, approximately; and silver-C has the smallest of 71.6 kDa. The average Mw of 72 kDa is fundamentally consistent with the deduction from SDS-PAGE (Yang et al. 2001). With respect to pI, five transferrin variants show three theoretic values in silver crucain carp: silver-D and silver-E have the lowest one of 5.67; silver-A and silver-B have a higher one of 5.74; while silver-C has the highest one of 5.90. As to white crucian carp, three transferrin variants display similar pI values of about 5.90.
Sequence Comparison and Polymorphism
Polymorphism analysis showed abundant variation among the 10 transferrin cDNA sequences of C. auratus, with the segregation site number of 248. Their deduced aa sequences were then aligned together with that of common carp transferrin for comparison (fig. 2). In contrast to the few substitutions detected among transferrin variants in other species (Welch and Langmead 1990; Carpenter and Broad 1993), a large number of variations were observed throughout the over-600 aa transferrin sequences, which are mainly composed of replacements and deletions. A number of conserved residues had been identified for transferrins from different species (Anderson et al. 1989), including two tyrosines (Tyr), a histidine (His), and an aspartic acid (Asp) within each transferrin half. However, significant substitutions happened in the N-terminal domain of C. auratus transferrin: two Tyr residues for iron binding were replaced by a His-101 (for silver-C, color-B1, and white-A) and an Asp-199 (glutamine for color-A1), respectively, whereas the arginine (Arg) or lysine (Lys) for anion binding was substituted by a serine residue (Ser-129). Furthermore, substantial substitutions also happened in conserved regions around these residues, with a faster accumulating tendency in those of the N-terminal domain than in those of the C-terminal (12 vs. 3) of transferrin variants of C. auratus.
Sequence identity among transferrin alleles of C. auratus varies greatly (table 4). The highest sequence homology was observed between silver-C and color-B1 (99.7% for the cDNA sequences and 99.5% for the aa sequences), and the lowest one was observed between silver-A and silver-C (91.4% for the cDNA sequences and 84.3% for the aa sequences). The average identity for transferrin cDNA and aa sequences were 93.9% and 89.1%, respectively.
Table 4 Percentage Sequence Identity of cDNA Sequences and Amino Acid Sequences Between Transferrin Alleles of C. auratus.
Positive Selection and Identification of Selected Sites
For simulation with various evolutionary models, a phylogenetic tree was first constructed for transferrin alleles of C. auratus based on their coding sequences using the maximum-likelihood algorithm (Felsenstein 1993). As shown in figure 3, these alleles did not show a tendency to form a single cluster within a same subspecies. Instead, partial transferrin alleles from the silver crucian carp showed an apparently closer phylogenetic relationship with those from the color crucian carp, e.g., silver-C and color-B1.
The free-ratio model of the PAML program assumed different dn/ds values among branches of the evolutionary tree. Of the 17 branches of ml tree based on transferrin sequences of C. auratus, seven showed the dn/ds value greater than 1.0 (fig. 3). The branches of larger dn/ds values did not show a converging distribution in certain allelic lineage. The log likelihood value under the free-ratio model was l1 = –4854.90. Meanwhile, under the one-ratio model, which assumed common dn/ds numbers of different branches parameter for the entire tree, the calculation led to l0 = –4866.39. Twice the log likelihood difference of two models was 2l = (l1 – l0) = 22.98. No remarkably significant difference was detected between these two models by 2 distribution (0.10 < P < 0.25; table 5) with the free degree of 16. There seemed to be no detectable difference in the rates of adaptive evolution among allelic lineages of C. auratus transferrin.
Table 5 Results of Likelihood Models.
Further calculation under the positive-selection model and the neutral model was practiced (table 5). Twice the log likelihood difference between these two models was 53.60, and results of the 2 distribution test provided a much better fit for the former model than for the latter (df = 2, P < 0.005). Furthermore, approximately 14% of the total codons were subjected to probable positive selection during their evolutionary history by the calculation, with an average value of 4.29. The sites, subject to no or strong constraint, occupied 31% and 55%, respectively (table 5). All these data presented substantial evidence for the existence of positive selection in the evolution of transferrin of C. auratus.
Empirical Bayes's approach was further applied to detect codons subjected to positive selection (Nielsen and Yang 1998). Altogether, 11 sites were concluded as subject to selection in transferrin sequences, for their posterior probabilities greater than 0.95. These included the codons Ile70, Gly86, Ala97, Ala142, Ile151, Arg254, Pro297, Pro330, Arg358, Asn411, and Gly428 of silver-A in figure 2. None of these aa residues were distributed within the conserved regions for iron ligand or anion binding of transferrin.
The correspondent residues of these selected sites were then identified in the spatial structure of rabbit transferrin (fig. 4; Hall et al. 2002), with the peptides for binding of bacterial protein similarly illustrated for comparison (Retzer, Yu, and Schryvers 1999). As a result, 10 selected sites found their correlates in rabbit transferrin sequence with the Ile151 left for the gap in sequence alignment (data not shown). The selected sites generally fall around the peptides for binding of bacterial proteins with respect to the salmonid transferrins (Ford 2001). However, only three of the 11 selected sites from C. auratus transferrin were located within these peptides. A test of 2 distribution implied no definite correlation between the distribution of these codons and the location of these peptides (0.25 < P < 0.75; df = 1). However, five of the selected sites, Val-60, Asn-76, Lys-92, Ser-248, and Ala-299, were located convergent in a plane internal in N-lobe. This plane exactly paralleled with the interface between the two domains—N1 and N2—of N-lobe with the iron atom vertically over it. Furthermore, four of the five remnant selected codons, Cys-137, Pro-332, Leu-365, and Lys-418, lay in a long and narrow region on the back of linkage between the two lobes of rabbit transferrin, while the last, Asn-435, was located without direct connection with any of the other selected sites. A common attribute of these five sites was that they were distributed on the surface of the transferrin molecule.
FIG. 4. Tubular model of the diferric rabbit serum transferrin molecule (Hall et al. 2002). The molecule is specially oriented for a clear view of the selected sites. Upper right, upper left, lower left, and lower right approximately illustrate the N1, N2, C1, and C2 domains of transferrin, respectively. Peptides to be contacted by bacterial transferrin-binding proteins (Retzer, Yu, and Schryvers 1999) are shown by dark tubules. Five protein ligands involved in binding of iron atom or the anion in the N-lobe of transferrin are indicated by white balls. The sites correspondent to those subject to positive selection in C. aurtus transferrin are illustrated by dark balls, and the codon number in sequence alignment is marked. The codon Ile146 did not have homologous resiude in the rabbit transferrin
Population Recombination Rates and Mosaic Sequences
By the coalescent estimator of the SITE program (Hey and Wakeley 1997), the population recombination rates of transferrin alleles in the C. auratus and the silver crucian carp were 0.0807 and 0.0403, respectively. Since most of C. auratus and silver crucian carp were tetraploidy and hexaploidy, respectively (Zhu and Jiang 1993), these numbers should be split by two and three and the ultimate values should be 0.0404 and 0.0135. Furthermore, the minimal set of recombination intervals (Hudson and Kaplan 1985) for transferrin of the C. auratus and the silver crucian carp were 26 and 10, respectively.
Ten coding sequences of C. auratus transferrin alleles were subsequently queried to those of the others using SIMPLOT2.5. Recombination was evidently observed for four of these sequences, including silver-A, silver-D, silver-E, and color-A1. No mosaicism was found among coding sequences of any of the transferrin alleles of white crucian carp. Furthermore, transferrin alleles of white crucian carp appeared to have an earlier divergence from those of the other C. auratus populations (fig. 3 and fig. 6); thus, the allele white-A was used as the outgroup in the further informative site analyses for determination of breakpoints among mosaic sequences (fig. 5). Recombination could be easily identified for the coding sequence of silver-A by comparing its sequence identity with that of silver-B and silver-D (fig. 5A). In the 5' and 3' ends of these sequences, most of the informative site trees supported a closer relationship between silver-A and silver-B. In the intermediate region, however, silver-A and silver-D shared more informative sites. Two recombination events were deduced between silver-B and silver-D in producing silver-A, probably through nucleotides 561–691 and 1275–1305 (P < 0.001 for both results). Similarly, silver-D could be considered as a recombination product of silver-E and silver-A with the breakpoints between nucleotides 881 and 936, 1341 and 1347, respectively (fig. 5B; P < 0.001 for both results). The mosaic sequences of silver-E allele were from silver-D and color-A1, with the breakpoint between nucleotides 753 and 897 (fig. 5C; P < 0.005). Also, a recombination event most likely happened among nucleotides 233–540 between silver-E and silver-D to produce color-A1 (data not shown, 0.005 < P < 0.01). Based on all these analyses, recombination seemed not to happen between the two regions of 936 to 1275 and 1347 to 1932 of transferrin coding sequences, and they were used for molecular clock calculation in the following analyses.
FIG. 5. Similarity plot for the transferrin alleles (A) silver-A, (B) silver-D, (C) and silver-E, with a window size of 400 bp and a step size of 10 bp. The two allelic sequences putative as the recombination parents were selected as majority (50%) consensus sequences of the subject allele to display the mosaicism. Dotted vertical lines indicate breakpoints identified by maximization of 2 as described in Materials and Methods, with numbers of informative sites shared by the subject sequence and the reference sequences indicated below in the color to that reference. P-values were calculated by using Fisher's exact test. Four-member trees consistent with these sites are shown on the left
Molecular Clock Tests and Estimation of Divergence Time
Two neighbor-joining trees were generated based on the middle region and 3' end of transferrin coding sequences, and the former is presented in figure 6. As shown in figure 6, the transferrin sequences from C. auratus and salmonids formed one distinct cluster. The salmonid transferrins could be divided into three major clusters exactly correspondent to its three taxonomic genera Salmo, Salvelinus, and Oncorhynchus. By contrast, the transferrin alleles from each subspecies of C. auratus did not form one monophyletic group. Partial alleles were closely related to each other within each C. auratus subspecies, i.e., white-B and white-C, silver-A and silver-D. However, transferrin alleles among populations could also share a close phylogenetic relationship, i.e., silver-C and color-B1, silver-E and color-A1. These data suggest that the divergence between transferrin alleles of C. auratus might occur before the geographic differentiation of its subspecies.
All the interior nodes showed the fundamentally identical evolutionary rate in the two-cluster test (Takezaki, Rzhetsky, and Nei 1995). However, two of the silver crucian carp transferrin alleles, silver-A and silver-D, evolved significantly faster than other copies at the 1% level in the branch-length test. Elimination of these two sequences favored a constant evolutionary rate among nodes and branches in both tests, and the U statistic for the two-cluster test was 19.0 (0.25 < P < 0.75; df = 17). A linearized tree was then constructed (fig. 7). The nine transferrin sequences of salmonids again formed three distinct clusters correspondent to their three genera, the divergence of which (node b) was used as a calibration point for estimation of the divergence time of C. auratus transferrin (node a). The fossil records of salmonid genera were 14 to 17 MYA for Salvelinus (unavailable unpublished data) and 14 MYA for Oncorhynchus (S. Ralph and G. R. Smith, unpublished data). The average heights of nodes b and a were 0.0184 ± 0.0045 and 0.0274 ± 0.0047, respectively. By referring to the smaller salmonid divergence of 14 Myr, polymorphism of C. auratus transferrin might have existed for 20.8 ± 3.6 Myr.
FIG. 7. Linearized tree constructed under the assumption of a molecular clock for the middle-region transferrin coding sequences of human, chicken, salmonids, and C. auratus. The divergence of two salmonid genera, Salvelinus and Oncorhynchus (node b), was used as the calibration point to deduce divergence of transferrin alleles of C. auratus (node a). The scale at the bottom indicates nucleotide substitutions per site
The apparently different phylogenetic relationship between transferrin alleles of C. auratus was observed in the neighbor-joining tree from 3'-end transferrin coding sequences (data not shown). After removal of the four salmonid (Atlantic salmon, brown trout, lake trout, and brook trout) and two C. auratus (silver-E and color-A1) sequences, a linearized tree was constructed (data not shown). Similarly, diversity of transferrin molecule in C. auratus was estimated to begin 15.5 ± 2.6 MYA by referring to the divergence of salmonid genera.
Discussion
Amino Acid Sequence Variation and Molecular Basis of C. auratus Transferrin Polymorphism
Molecular mechanism of transferrin variations has been studied in many species from fishes to mammals (Hershberger 1970; Welch and Langmead 1990; Carpenter and Broad 1993; Laurent and Rodellar 2001). The genetic basis of transferrin polymorphism has been preliminarily discussed in C. a. auratus by using a simple model of color crucian carp (Yang, Zhou, and Gui 2004). A hypothesis was put forward that mobility of transferrin variants was negatively relevant to their Mw on SDS-PAGE gel and negatively relevant to their pI on native PAGE gel.
In this study, more cDNA sequences of transferrin alleles have been cloned from the silver crucian carp and white crucian carp (table 2). The deduced transferrin variants showed variable aa sequences and distinct computed physicochemical properties (fig. 2; table 3). By using the previous hypothesis on polymorphism of transferrin in color crucian carp (Yang, Zhou, and Gui 2004), a schematic graphic has been given to interpret the more complicated transferrin phenotypes of the silver crucian carp (fig. 8). With respect to the transferrin phenotypes on 8% PAGE gel, the distinct transferrin bands have been explained as encoded by three transferrin alleles of Tfa, Tfb, and Tfc (Yang et al. 2001). However, five silver crucian carp transferrin variants have been identified in this study by their variable cDNA sequences (table 2). As to their theoretical Mw, the condition is: silver-D > (silver-A = silver-B = silver-E) > silver-C (table 3). Hypothetically, three bands of different mobility were to emerge on SDS-PAGE gel. However, only two bands could be distinguished for these silver crucian carp clones on 12% SDS-PAGE gel (Yang et al. 2001). Most probably silver-D was not distinguished from silver-A, -B, and -E for the limited discrimination of PAGE, and silver-C encoded the faster band shared by clone D and clone A (fig. 8A). Meanwhile, five deduced transferrin variants show three pI values in silver crucian carp; they were aligned as: (silver-D = silver-E) < (silver-A = silver-B) < silver-C (table 3). Three transferrin bands were thus predicted for silver crucian carp on 8% PAGE, which corresponded to, respectively, the three transferrin alleles assumed previously (Yang et al. 2001). Namely, band a had the largest anodal migration and was encoded by silver-D or silver-E, band b ran at the intermediate speed and was encoded by silver-A/B, and the slowest band, c, was the product of allele silver-C (fig. 8B).
FIG. 8. The schematic transferrin phenotypes of silver crucian carp under the hypothesis that mobility of transferrin variants was determined by their Mw in SDS-PAGE (A) or by their pI in native PAGE (B). Transferrin bands are shown by continuous lines except for the discontinuous silver-A in clone A of silver crucian carp for its non-detection in PAGE. The transferrin genotypes of silver crucian carp clones (table 2) are expressed such that the transferrin alleles are listed in order of larger electrophoretic migration. The lowercase letters a, b, and c show the transferrin alleles previously assumed (Yang et al., 2001)
Transferrin allele silver-A was detected from all the three silver crucian carp clones by PCR (table 2). Its encoded band b was observed for clones D and F of silver crucian carp on 8% PAGE gel; however, it was not detected for clone A (Yang et al. 2001; indicated by discontinuous lines in fig. 8B). Plausibly deleterious variation happened in gene of silver-A and the abnormal expression resulted in its non-detection on PAGE gel, considering that the silver crucian carp was a unisexual and polyploid animal and tended to accumulate more mutations in its genome (Muller 1964; Risinger and Larhammar 1993).
Transferrin polymorphism was also observed between the two white crucian carp specimens (data not shown). Considerable variations were observed among the three transferrin alleles identified from the white crucian carp (fig. 2), but their genetic relationships with the transferrin variants have not been determined. Particularly, one potential N-glycosylation site was identified for transferrin sequences of white-A and white-B (fig. 2). Consequently, a more complicated condition has to be considered in the further discussion of genetic basis of transferrin polymorphism in the C. auratus.
Identification of Selected Transferrin Sites and Their Function Significance
Codons subject to positive selection had been identified in salmonid transferrins, and their homologous sites displayed the directional distribution in the peptides for binding of bacterial proteins in the bovine lactoferrin (Ford 2001). Polymorphic residues also showed a tendency to locate on external parts of transferrin variants in horse (Carpenter and Broad 1993). However, testing of 2 distribution refuted any association of the sites homologous to those selected in C. auratus transferrin with the peptides for bacterial binding in rabbit transferrin. In other words, the rapid evolution of transferrin alleles of C. auratus did not seem to be driven by the competition of iron by the pathogenic proteins.
In contrast to their irregular distribution in the primary structure of transferrin, however, five of the 11 homologous selected residues converged in a plane closely parallel to the cleft between the N1 and N2 domains (fig. 8). Studies of the crystal structure of human lactoferrin showed the probable condition that iron and anion were bonded to transferrin within the interdomain cleft (Anderson et al. 1989). The iron atom is coordinated to four protein ligands and the specific CO32- bound to iron as a bidentate bond. A few polar residues around the iron-binding cavity form hydrogen bonds with the anion or the iron, and more ones provided a hydrophilic environment for transferrin to bond to iron or anion. The five selected sites converged in the interdomain of the N-lobe of C. auratus transferrin could be replaced as the aa of Glu, Asp, Arg, His, and Cys (see fig. 2 for the possible replacements). These polar residues had potentiality to form hydrogen bonds with those directly bonded to iron and anion, and their situation in providing a hydrophilic cavity could be expected. Consequently, variation of these five selection sites might indirectly influence the capability of C. auratus transferrin to bind to or release iron. The other five correspondent selected sites appeared on the surface of rabbit transferrin and were not even near the interdomain cleft for binding of ions. However, these five residues were all located at linkages of different secondary structures (fig. 9); variations of them might result in the opening or closing of interdomain clefts or half molecules to different extents and ultimately might have an effect on the binding or releasing of iron (Anderson et al. 1989).
FIG. 9. Secondary structure of the diferric rabbit serum transferrin molecule (Hall et al. 2002). Helixes and strands are respectively indicated by cylinders and arrows, respectively. Residues with the probable function with iron or anion and those homologous to the selected sites in transferrin of C. auratus are indicated by white balls and black balls, respectively
These analyses implied that evolution of the selected sites in transferrin of C. auratus could probably interfere with its function in the uptake of iron. Iron is an essential element for the body because it is necessary for normal conformation and function of some iron-carrying proteins, such as hemoglobin and myoglobin. As a result, metabolisms of iron and oxygen are associated closely in the organism. For instance, expression of transferrin and its acceptor were simultaneously activated in the organism in the condition of hypoxia (Rolfs et al. 1997; Tacchini et al. 1999). The genus Carassius has long been known for its capability being of anoxia-tolerant (Lutz, Nilsson, and Perez-Pinzon 1996). Furthermore, C. auratus inhabits a vast territory of the Eurasia continent and has to experience rough environments that include drought or cold, both coupled with hypoxia. Consequently, diversification of C. auratus transferrin was plausibly firmly driven by these extreme environments, and the resulted abundant transferrin genotypes might provide better survival of C. auratus. Presumably, research on anoxia-tolerance of C. auratus individuals with distinct transferrin phenotypes might provide further proof of this hypothesis.
Phylogenetic Relationship and Antique Origin of C. auratus Transferrin Alleles
Two phylogenetic trees were constructed based on different regions of coding sequences of C. auratus transferrin (fig. 6; data not shown). Both phylogenies inferred the closer phylogenetic relationship between transferrins of silver crucian carp and color crucian carp, whereas the transferrin alleles of white crucian carp tended to form a monophylogeny. Remarkably, the highest sequence homology was found between the transferrin alleles of silver-C and color-B1, which were from two C. auratus subspecies with differentiation in both morph and habitat (Luo and Yue 2000). Besides these phylogenetic analyses, recombination events were occasionally observed between transferrin alleles of silver crucian carp and color crucian carp (fig. 5C). These data inferred the significant influence of geographic distribution on differentiation of three C. auratus populations. In other words, the white crucian carp was subject to more complete genetic isolation from other C. auratus populations on the separate Japanese islands. By contrast, silver crucian carp and color crucian carp inhabited the Asian continent and partially shared their gene pool. Hybridization introgression could occasionally happen between these two populations with recombination observed.
Molecular clock calculation inferred an evolutionary history of up to 15–20 Myr for the C. auratus transferrin alleles. The tetraploidization event in cyprinids was deduced to occur 16–22 MYA, estimated by divergence of the synaptosome-associated protein loci (Risinger and Larhammar 1993), and 50–70 MYA based on hemoglobin data (Powers and Edmundson 1972). Although it was unclear if these transferrin gene lineages emerged as the direct result of the polyploidization event, it might exert significant influence on their further diversification (see the next section). Some data (Yang, Zhou, and Gui 2004) has supported the allelic origin of the transferrin copies. They are considered as alleles in the following sections. However, further studies on their genomic localization and gene structure are necessary to clarify their origin.
Evolutionary Mechanism of High Transferrin Polymorphism in C. auratus
DNA polymorphism of transferrin has been studied in mammals (Carpenter and Broad 1993; Laurent and Rodellar 2001) and fishes (Ford 2000; Antunes et al. 2002). About 20 nucleotide sites were identified as variable among transferrin alleles, and fewer aa replacements were detected within these species. However, up to 248 variable nucleotide sites have been identified among the 10 transferrin alleles of C. auratus in this study. The average sequence identity of 89.5% among transferrin variants was even comparable to that among salmonid species (Lee et al. 1998). Then how has this redundant DNA polymorphism of transferrin happened?
The ancient allelic lineages of C. auratus might give the first interpretation. It was known that speciation of the C. auratus was not too long before 500 thousand years ago, when progenitor white crucian carp emerged (Murakami, Matsuba, and Fujitani 2001). Consequently, DNA polymorphism of C. auratus transferrin most likely derived earlier than the species C. auratus itself and was a transspecies polymorphism (Klein 1986). Furthermore, generally these allelic lineages did not diverge earlier than the tetraploidization event of cyprinids. The gene duplication coupled by polyploidization might provide doubled or even tripled (for the silver crucian carp; Zhu and Jiang 1993) genetic units for further diversification, such as recombination and selection.
Detection of codons subject to positive selection presented the second explanation. Most of the duplicated genes were silenced within a few Myr, and even the few survivors would probably experience strong purifying selection (Lynch and Conery 2000). However, positive selection could still been observed for some genes and played the important role in diversification of them for its accelerative fixation of replacement mutations (Hughes and Nei 1988; Yang 1998). Natural selection exerted on partial codons of C. auratus transferrin could elevate their aa substitutions and cause significant effects on its allelic diversification.
Furthermore, recombination was thought to play a role in the generation of novel alleles at various human major histocompatibility complex loci (Carrington 1999). It probably played the similar role in the generation of genetic variations in C. auratus transferrin, because four of its alleles showed the apparent recombination origin (fig. 4). Remarkably, sex and recombination accelerate the accumulation of beneficial mutation and reduce accumulation of harmful mutation by increasing realistic selection strength from background trapping (Rice and Chippindale 2001). Therefore, recombination and selection might interact with each other in increasing genetic variation of C. auratus transferrin.
The antique allelic lineages of C. auratus transferrin strongly implied a balancing selection mechanism in the maintenance of their polymorphism, as neutral alleles could only persist in a population for limited generations after their fixation (Takahata and Nei 1990).
Two forms of balancing selection might then happen, overdominant or frequency-dependent, with the former preferable for heterozygotes and the latter tending to retain the genotype of lower allele frequencies in the population. High transferrin heterozygosity has been observed to be prevalent in different fishes (Kirpichnikov 1981). Zhang (1996) investigated the transferrin polymorphism in a wild C. auratus population (C. auratus auratus). Most of the individuals (24 vs. 26) showed the heterozygous transferrin phenotypes. Furthermore, 18 of the 28 individuals of the color crucian carp were transferrin heterozygotes in our studies (Yang, Zhou, and Gui 2004), as well as were all the silver crucian carp clones (Yang et al. 2001). Collectively, these data suggested a heterozygosity not less than 50% in the C. auratus population. Furthermore, a significant characteristic of overdominant selection might be the detection of positive selection for its existence as a strong drive for novel alleles with substantial variations and higher heterzygosity in the population (Li 1997, pp. 237–268). Positive selection was evidently confirmed for transferrin alleles of C. auratus, and the sites subject to selection were also identified. Based on these results, overdominance might be the more possible selective mechanism for long persistence of transferrin polymorphism in the population. However, direct data of comparison of fitness among transferrin phenotypes and investigation of allele frequency distribution were requisite to test this type of balancing selection.
The silver crucian carp is a special C. auratus population with complicated evolutionary backgrounds, which might exert significant influence on the selective mechanism of its transferrin. With the clonal reproductive mode without recombination, mutation tends to rapidly fix in the clone. Most of the mutations are neutral in expectation. However, the silver crucian carp clone with a new mutation is most probably a heterozygote for the asymmetry of mutation, which has the higher fitness under both types of balancing selection. Subsequently, this clone will propagate quickly without the "cost of males" (Smith 1975) and will completely occupy its niche. Once the environment has changed, fitness of this clone will lower to moderate or poor. The old transferrin allele will be lost easily in genetic drift or even purged by negative selection because it cannot participate in the creation of a novel genotype for absence of recombination. Apparently, positive selection can be easily observed in the clonal population of silver crucian carp, but the long persistence of transferrin polymorphism seems impossible.
Interestingly, silver crucian carp is never an asexual population absolutely separate in the nature. A minority ratio of males was found in the silver crucian carp populations and often lived with its bisexual related populations, such as C. a. auratus, Carassius carassius, and common carp (Cherfas 1981; Fan and Shen 1990). Significantly, sexual reproduction and recombination have been observed among silver crucian carp clones (Zhou, Wang, and Gui 2000; Zhou and Gui 2001). In the current study, we have also revealed the recombination events among their transferrin alleles (fig. 5).
Vrijenhoek's frozen niche variation model (1979) could be used to explore the realistic balancing selection mechanism in the transferrin of silver crucian carp. As is the condition in other unisexual vertebrates, the silver crucian carp probably has a hybridization origin (Cherfas 1981; Yang et al. 2001). The silver crucian carp clones continually capture and freeze genetic variation from their sexual ancestors, and multiple heterozygotic transferrin genotypes are therefore fixed. Selection among clones favors specialized genotypes having minimal niche overlap with other clones and the sexual ancestors (the marginal overdominance; Wallace 1968). Furthermore, the novel transferrin alleles and genotypes can come from hybridization between silver crucian carp clones or silver crucian carp and their related bisexual populations, which are fixed again in the form of clone (Zhou, Wang, and Gui 2000).
Presumably, the versatile balancing selection type enables silver crucian carp to survive in complicated environments. Stronger positive selection can be expected in the silver crucian carp clones for the rapid fixation of advantageous mutations in gynogenetic clones of silver crucian carp (fig. 3, the silver-A/silver-B cluster). Recombination can not only create abundant genetic variations, but it can also remove the accumulating mutational load in the asexual population (Rice 1998). Significantly, novel alleles can be introduced into the silver crucian carp clones for their hybridization origin, and the obsolete alleles may participate in the creation of new transferrin phenotypes by occasional sexual reproduction, which enables the persistence of ancient allelic lineages in the population. Therefore, owing to its particular evolutionary backgrounds, the silver crucian carp might possess a distinctive balancing selection mechanism.
FIG. 6. Neighbor-joining tree for transferrins of chicken, salmonids and C. auratus, with human assumed as the outgroup. The branch lengths (see scale at bottom) are measured in terms of the number of nucleotide substitutions per site based on the middle-region coding sequences of transferrins (332 nucleotides). Numbers given to interior branches are the bootstrap values higher than 50% with 1,000 replicate resamplings (Kumar et al. 2001). Three salmonid genera, Salmo, Salvelinus, and Oncorhynchus, are indicated by brackets. Asterisks (**) indicate the transferrin alleles evolving significantly faster at the 1% level in the branch-length test (Takezaki, Rzhetsky, and Nei 1995)
Acknowledgements
This work was supported by the National Natural Science Foundation of China (grant numbers 30200028, 30130240, and 30123004), the Chinese Academy of Sciences (KSCX2-SW-303, KSCX2-1–05), the Institute of Hydrobiology of the Chinese Academy of Sciences (220309), and the European Commission (contract no. ICA4-CT-2001–10024).
Literature Cited
Anderson, B. F., H. M. Baker, G. E. Norris, D. W. Rice, and E. N. Baker. 1989. Structure of human lactoferrin: crystallographic structure analysis and refinement at 2.8 ? resolution. J. Mol. Biol. 209:711-734.
Antunes, A., A. R. Templeton, R. Guyomard, and P. Alexandrino. 2002. The role of nuclear genes in intraspecific evolutionary inference: genealogy of the transferring gene in the brown trout. Mol. Biol. Evol. 19:1272-1287.
Arbogast, B. S., S. V. Edwards, J. Wakeley, P. Beerli, and J. B. Slowinski1. 2002. Estimating divergence times from molecular data on phylogenetic and population genetic time scales. Annu. Rev. Ecol. Syst. 33:707-740.
Baldwin, G. S. 1993. Comparison of transferrin sequences from different species. Comp. Biochem. Physiol. 106B:203-218.
Carpenter, M. A., and T. E. Broad. 1993. Polymorphism in the coding sequence of the horse transferrin gene. Genome 37:157-164.
Carrington, M. 1999. Recombination within the human MHC. Immunol. Rev. 167:245-256.
Cherfas, N. B. 1981. Gynogenesis in fishes. Pp. 255–273 in V. S. Kirpichnikov, ed. Genetic bases of fish selection. Springer-Verlag, Berlin.
Fan, Z., and J. Shen. 1990. Studies on the evolution of bisexual reproduction in crucian carp (Carassius auratus gibelio Bloch). Aquaculture 84:235-244.
Felsenstein, J. 1993. PHYLIP (phylogeny inference package). Version 3.5c. Distributed by the author, Department of Genetics, University of Washington, Seattle.
Ford, M. J. 2000. Effects of natural selection on patterns of DNA sequence variation at the transferrin, somatolactin, and p53 genes within and among chinook salmon (Oncorhynchus tshawytscha) populations. Mol. Ecol. 9:843-855.
Ford, M. J. 2001. Molecular evolution of transferrin: evidence for positive selection in salmonids. Mol. Biol. Evol. 18:639-647.
Frelinger, J. A. 1972. The maintenance of transferrin polymorphism in pigeons. Proc. Natl. Acad. Sci. USA 69:326-329.
Gui, J. F. 1996. A unique study system: gynogenetic fish Carassius auratus gibelio. Science Foundation of China 4:44-46.
Gui, J. F., S.C. Liang, L. F. Zhu, and Y. G. Jiang. 1993. Discovery of two different reproductive development modes of the eggs of artificial multiple tetraploid allogynogenetic silver crucian carp. Chinese Sci. Bull. 38:332-337.
Hall, D. R., J. M. Hadden, G. A. Leonard, S. Bailey, M. Neu, M. Winn, and P. F. Lindley. 2002. The crystal and molecular structures of diferric porcine and rabbit serum transferrins at resolutions of 2.15 and 2.60 ?, respectively. Acta Crystallogr. D Biol. Crystallogr. 58:70-80.
Hershberger, C. L., J. L. Larson, and B. Arnold, et al. (12 co-authors). 1991. A cloned gene for human transferrin. Ann. N.Y. Acad. Sci. 646:140-154.
Hershberger, W. K. 1970. Some physicochemical properties of transferrins in brook trout. Trans. Amer. Fish. Soc. 1970:207-218.
Hey, J., and J. Wakeley. 1997. A coalescent estimator of the population recombination rate. Genetics 145:833-846.
Hudson, R. R., and N. L. Kaplan. 1985. Statistical properties of the number of recombination events in the history of a sample of DNA sequences. Genetics 111:147-164.
Hughes, A. L., and M. Nei. 1988. Pattern of nucleotide substitution at major histocompatibility complex class I loci reveals overdominant selection. Nature 335:167-170.
Jeltsch, J. M., and P. Chambon. 1982. The complete nucleotide sequence of the chicken ovotransferrin mRNA. Eur. J. Biochem. 122:291-295.
Kimura, M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. J. Mol. Evol. 16:111-120.
Kirpichnikov, V. S. 1981. The genetics of nonenzymatic proteins in fishes. Pp. 162–172 in V. S. Kirpichnikov, ed. Genetic bases of fish selection. Springer-Verlag, Berlin.
Klein, J. 1986. Natural history of the major histocompatibility complex. John Wiley, New York.
Kumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001. MEGA2: molecular evolutionary genetics analysis software. Bioinformatics 17:1244-1245.
Kvingedal, A. M., K. A. Rorvik, and P. Alestrom. 1993. Cloning and characterization of Atlantic salmon (Salmo salar) serum transferrin cDNA. Mol. Mar. Biol. Biotechnol. 2:233-238.
Laurent, P., and C. Rodellar. 2001. Characterization of a single nucleotide polymorphism in the coding sequence of the bovine transferrin gene. Mut. Res. Genomics 458:1-5.
Lee, J. Y., T. Tada, I. Hirono, and T. Aoki. 1998. Molecular cloning and evolution of transferrin cDNAs in salmonids. Mol. Mar. Biol. Biotech. 7:287-293.
Lee, J.-Y., N. Tange, H. Yamashita, I. Hirono, and T. Aoki. 1995. Cloning and characterization of transferrin cDNA from coho salmon (Oncorhynchus kisutch). Fish Pathol. 30:271-277.
Li, W. H. 1997. Molecular evolution. Sinauer Associates, Sunderland, Mass.
Luo, Y. L., and P. Q. Yue. 2000. Cyprininae. Pp. 427–433 in P. Q. Yue et al, eds. Fauna sinica, Osteichthyes, Cypriniformes III. Science Press, Beijing.
Lutz, P. L., G. E. Nilsson, and M. A. Perez-Pinzon. 1996. Anoxia tolerant animals from a neurobiological perspective. Comp Biochem Physiol 113B:3-13.
Lynch, M., and J. S. Conery. 2000. The evolutionary fate and consequences of duplicate genes. Science 290:1151-1155.
Muller, H. J. 1964. The relation of recombination to mutational advance. Mutat. Res. 1:2-9.
Murakami, M., C. Matsuba, and H. Fujitani. 2001. The maternal origins of the triploid ginbuna (Carassius auratus langsdorfi): phylogenetic relationships within the C. auratus taxa by partial mitochondrial D-loop sequencing. Genes Genet. Syst. 76:25-32.
Nielsen, R., and Z. Yang. 1998. Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene. Genetics 148:929-936.
Ohno, S. 1970. Evolution by gene duplication. Springer, Berlin.
Powers, D. A., and A. B. Edmundson. 1972. Multiple hemoglobins of catostomid fish. II. The amino acid sequence of the major chain from Catostomus clarkii hemoglobins. J. Biol. Chem. 247:6694-6707.
Retzer, M. D., R. H. Yu, and A. B. Schryvers. 1999. Identification of sequences in human transferrin that bind to the bacterial receptor protein, transferrin-binding protein B. Mol. Microbiol. 32:111-121.
Rice, W. R. 1998. Requisite mutational load, pathway epistasis, and deterministic mutation accumulation in sexual versus asexual populations. Genetica 102/103:71-81.
Rice, W. R., and A. K. Chippindale. 2001. Sexual recombination and the power of natural selection. Science 294:555-559.
Risinger, C., and D. Larhammer. 1993. Multiple loci for synapse protein SNAP-25 in the tetraploid goldfish. Proc. Natl. Acad. Sci. USA 90:10598-10602.
Robertson, D. L., B. H. Hahn, and P. M. Sharp. 1995. Recombination in AIDS viruses. J. Mol. Evol. 40:249-259.
Rolfs, A., I. Kvietikova, M. Gassmann, and R. H. Wenger. 1997. Oxygen-regulated transferrin expression is mediated by hypoxia-inducible factor-1. J. Biol. Chem. 272:20055-20062.
Smith, J. M. 1975. Theory of evolution. 3rd edition. Harmondsworth: Penguin.
Tacchini, L., L. Bianchi, A. Bernelli-Zazzera, and G. Cairo. 1999. Transferrin receptor induction by hypoxia: HIF-1-mediated transcriptional activation and cell-specific post-transcriptional regulation. J. Biol. Chem. 274:24142-24146.
Takahata, N., and M. Nei. 1990. Allelic genealogy under overdominant and frequency-dependent selection and polymorphism of major histocompatibility complex loci. Genetics 124:967-978.
Takezaki, N., A. Rzhetsky, and M. Nei. 1995. Phylogenetic test of the molecular clock and linearized trees. Mol. Biol. Evol. 12:823-833.
Tange, N., J.-Y. Lee, N. Mikawa, I. Hirono, and T. Aoki. 1997. Cloning and characterization of transferrin cDNA and rapid detection of transferrin gene polymorphism in rainbow trout (Oncorhynchus mykiss). Mol. Mar. Biol. Biotechnol. 6:354-359.
Thompson, J. D., T. J. Gibson, F. Plewniak, F. Jeanmoughin, and D. G. Higgins. 1997. The ClustalX windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 24:4876-4882.
Vrijenhoek, R. C. 1979. Factors affecting clonal diversity and coexistence. Am. Zool. 19:787-797.
Wallace, B. 1968. Topic in population genetics. W. W. Norton, Inc., New York.
Wei, W. H., J. Zhang, Y. B. Zhang, L. Zhou, and J. F. Gui. 2003. Genetic heterogeneity and ploidy level analysis among different gynogenetic clones of polyploid gibel carp. Cytometry 56:46-52.
Welch, S., and L. Langmead. 1990. A comparison of the structure and properties of normal human transferrin and a genetic variant of human transferrin. Int. J. Biochem. 22:275-282.
Williams, J. 1968. A comparison of glycopeptides from the ovotransferrin and serum transferrin of the hen. Biochem. J. 108:57-67.
Wilson, M. R., D. Middleton, and G. W. Warr. 1988. Immunoglobulin heavy chain variable region gene evolution: structure and family relationships of two genes and a pseudogene in a teleost fish. Proc. Natl. Acad. Sci. USA 85:1566-1570.
Yang, L., S. T. Yang, X. H. Wei, and J. F. Gui. 2001. Genetic diversity among different clones of the gynogenetic silver crucian carp, Carassius auratus gibelio, revealed by transferrin and isozyme markers. Biochem. Genet. 39:213-225.
Yang, L., L. Zhou, and J. F. Gui. 2004. Molecular basis of transferrin polymorphism in goldfish (Carassius auratus). Genetica (in press).
Yang, Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. CABIOS 13:555-556.
Yang, Z. 1998. Likelihood ratio tests for detecting positive selection and application to primate lysozyme evolution. Mol. Biol. Evol. 15:568-573.
Yi, M. S., Y. Q. Li, J. D. Liu, L. Zhou, Q. X. Yu, and J. F. Gui. 2003. Molecular cytogenetic detection of paternal chromosome fragments in allogynogenetic gibel carp, Carassius auratus gibelio. Bloch. Chromosome Res. 11:665-671.
Yu, X. J., T. Zhou, Y. C. Li, K. Li, and M. Zhou. 1989. Chromosomes of Chinese fresh-water fishes. Science Press, Beijing.
Zhang, H. 1996. Studies on the genetic population structure of aquatic animals. Doctoral dissertation, Chinese Academy of Sciences.
Zhou, L., and J. F. Gui. 2001. Studies on genetic diversity in gonochoristic offspring produced from mating between two different gonogenetic clones of silver crucian carp. Acta Biologiae Experimentalis Sinica 34:169-176.
Zhou, L., and J. F. Gui. 2002. Karyotypic diversity in polyploid gibel carp, Carassius auratus gibelio Bloch. Genetica 115:223-232.
Zhou, L., Y. Wang, and J. F. Gui. 2000. Genetic evidence for gonochoristic reproduction in gynogenetic silver crucian carp (Carassius auratus gibelio Bloch) as revealed by RAPD assays. J. Mol. Evol. 51:498-506.
Zhu, L. F., and Y. G. Jiang. 1993. A comparative study of the biological characters of gynogenetic clones of silver crucian carp (Carassius auratus gibelio). Acta Hydrobiologica sinica 17:112-120.(Lin Yang*, and Jian-Fang )