Genomic and Transcriptional Profiles of Kelch-like (klhl) Gene Family in Polyploid Carassius Complex

Abstract

Genome duplication supplies raw genetic materials and has been thought to be essential for evolutionary innovation and ecological adaptation. Here, we select Kelch-like (klhl) genes to study the evolution of the duplicated genes in the polyploid Carassius complex, including amphidiploid C. auratus and amphitriploid C. gibelio. Phylogenetic, chromosomal location and read coverage analyses indicate that most of Carassius klhl genes exhibit a 2:1 relationship with zebrafish orthologs and confirm two rounds of polyploidy, an allotetraploidy followed by an autotriploidy, occurred during Carassius evolution. The lineage-specific expansion and biased retention/loss of klhl genes are also found in Carassius. Transcriptome analyses across eight adult tissues and seven embryogenesis stages reveal varied expression dominance and divergence between the two species. The expression of klhls in response to Carassius herpesvirus 2 infection shows different expression changes corresponding to distinct herpesvirus resistances in three C. gibelio gynogenetic clones. Finally, we find that most C. gibelio klhl genes possess three alleles except eight genes that have lost one or two alleles due to genome rearrangement. The allele expression bias is prosperous for Cgklhl genes and varies during embryogenesis owning to the sequential expression manner of the alleles. The current study provides global insights into the genomic and transcriptional evolution of duplicated genes in a given superfamily resulting from multiple rounds of polyploidization.

1. Introduction

The Kelch-like (klhl) gene family encodes a class of substrate adapter proteins for Cullin-E3 ubiquitination and is characterized by a broad-complex, tramtrack, bricá-brac/poxvirus and zinc finger (BTB/POZ) domain, a BACK domain, and five to six Kelch motifs. A total of 42 KLHL genes have been identified in the human genome [1]. They share similar secondary structures, but their primary sequences have little homology. Most KLHL proteins can interact with Cullin3 to form Cullin-E3 ubiquitin ligase complex via BTB/POZ domain and act as adapters to recruit substrates for ubiquitination through Kelch motifs. Although functions of many KLHL proteins remain unexplored, they have been considered crucial regulators in a variety of fundamental cellular processes [2,3,4]. In recent years, increasing evidence has shown that loss-of-function mutations and dysregulated expression of KLHLs are closely associated with cancers, hereditary diseases, and chronic disorders, resulting in KLHLs attracting considerable attention as potential therapeutic targets [4,5,6,7,8]. Up to now, research and information on the fish klhl gene family are scarce. Only a few members, such as keap1 (kelch-like ECH-associated protein 1, also known as klhl19), klhl15, klhl31, and klhl41 have been identified from several fish species, and the majority of these studies focused on the expression of keap1 and its substrate nuclear factor erythroid 2-related factor 2 in response to dietary nutrients or environmental contaminants to assess the antioxidant activities.

Besides the two rounds of whole genome duplication (WGD) that occurred at the root of vertebrates (1R and 2R), it has been proposed that a teleost-specific third WGD (Ts3R) occurs in the ancestor of teleost [9,10,11]. Additionally, polyploidization has independently and repeatedly occurred in several taxonomic fish orders [9,10,11]. For example, about 400 polyploid species across over 30 genera have been discovered in Cyprininae [12]. WGD has been considered a main driving force of speciation owing to its profound impacts on evolutionary innovation and ecological adaptation [13,14,15,16]. One consequence of WGD is expression alteration in entire networks involved in hundreds to thousands of duplicated genes, termed “expression evolution” or “regulatory evolution” [17]. Although the expression evolution after WGD has been well documented in paleopolyploids [13,18], we still have a limited understanding of young polyploids with complex genomes.

Carassius is an extraordinary genus, including multiple-ploidy organisms that reproduce by different reproduction modes [13,18]. Of them, sexual C. auratus (crucian carp/goldfish) is an amphidiploid (also called allotetraploid) (2n = 4x = 100, AABB), and gynogenetic C. gibelio (gibel carp) is an amphitriploid (previously called allo-/autohexaploid) (3n = 6x = ~150, AAABBB) that contains two subgenomes, and each subgenome has three homologs (Figure 1A) [19,20,21]. Carassius is distributed widely across the Eurasian continent [22,23,24] and is one of the most important aquaculture species in China [25,26,27]. In addition to Ts3R, it has been speculated that two extra rounds of polyploidy, an early allotetraploidy 10.17 million years ago (Mya) and a later autotriploidy 0.82 Mya, occurred during Carassius evolution (Figure 1A) [19]. After autotriploidy, C. gibelio evolved unique multiple reproduction modes and sex determination mechanisms [28,29]. Therefore, the Carassius complex provides a valuable chance to inspect the fates of duplicated genes and their impacts on evolutionary innovation and ecological adaptation. In this study, we explored the evolutionary trajectories of the klhl gene family in Carassius and investigated their expression panorama after experiencing multiple rounds of polyploidy.

2. Results

2.1. Genome-Wide Identification of klhl Genes

To explore the evolutionary origin of the klhl gene family, we retrieved klhl genes via a hierarchical strategy from the assemblies of C. auratus and C. gibelio, as well as from other 11 animals (Figure 1B). None of the klhl genes was identified from the genome of Cephalochordata amphioxus (Branchiostoma belcheri), a living invertebrate close to vertebrates [30]. Sea lamprey (Petromyzon marinus), a representative of the most ancient vertebrate [31,32], had the fewest klhl genes (20), suggesting that klhl genes would emerge during early vertebrate evolution.

Spotted gar (Lepisosteus oculatus), a teleost sister lineage that did not experience the Ts3R [33], had 44 klhl genes, whereas six teleosts analyzed in this study all possessed more klhl genes. For example, 67 klhl genes were identified in the zebrafish (Danio rerio) genome, of which 20 members had at least two duplicates that would be generated by either Ts3R or small-scale duplication. As paleo-tetraploids [34,35], rainbow trout (Oncorhynchus mykiss), and Atlantic salmon (Salmo salar) possessed 67 and 73 klhl genes in their genomes, respectively. Although the numbers were comparable to those of zebrafish, some klhl genes still had one more copy in these two Salmonidae fishes (Table S1). In contrast, 96 and 98 klhl genes were identified in C. auratus and C. gibelio, respectively, and most of them had one more duplicate (Table S1). It was confirmed that the two assemblies are both composed of one haplotype of two subgenomes (AB). Although possessing similar gene numbers, minor allele frequencies of single nucleotide polymorphism (SNP) sites in klhl genes were around 0.34 (approximately 1/3) in C. gibelio and around 0.48 (approximately 1/2) in C. auratus (Figure 1C). These data indicate that C. gibelio usually has three alleles, and C. auratus has two alleles for each klhl gene, consistent with their polyploidy (Figure 1A). In addition, klhl1 was likely lost in the ancestor of teleost, and klhl43 was likely a new member that emerged in ray-finned fishes after diverging from tetrapods.

2.2. Evolution of klhl Gene Family in Carassius

To clarify orthologous relationships, we constructed an ML phylogenetic tree containing all Klhl members identified from zebrafish, C. auratus, and C. gibelio assemblies. Of a total of 55 gene branches (Figure 2A), 28 klhl genes showed a ratio of 1:2:2 among zebrafish, C. auratus, and C. gibelio (e.g., klhl33 in group Ⅲ, klhl6 in group Ⅳ, and klhl4 in group V). Gene expansions in a special fish were also observed, including 10 genes in zebrafish (e.g., klhl2 in group V and klhl9 in group III), one in C. auratus (klhl30 in group Ⅳ) and two in C. gibelio (klhl8 and klhl39b in group V). The rest were concerned with gene loss. For example, klhl27 (group V) and klhl34 (group III) were missing in zebrafish, and 15 genes had only one duplicate in both Carassius species (e.g., klhl3, klhl7, and keap1b in group V). However, we did not find any special gene loss only in C. auratus or in C. gibelio.

To clearly present evolutionary trajectories of the klhl gene family in Carassius, we explored their chromosome localization (Tables S2 and S3) and classified them into four categories (Table S4). Consistently, 35 out of 55 klhl genes (category I) had two duplicates that just resided at a pair of homeologous chromosomes, both in C. auratus and C. gibelio, forming a 2:2:1 relationship with the zebrafish homolog (Figure 2B, red lines). Therefore, these two klhl genes were a pair of homeologs named by adding ‘-A’/’-B’ after the symbol of their zebrafish orthologue (e.g., klhl1a-A and klhl1a-B). Besides possessing two homeologous duplicates, klhl30 and klhl39b (category II) appeared to have more gene copies (Figure 2B, blue lines). These extra gene expansions only occurred in one species and may come from small-scale duplication. The third category contained 14 klhl genes that displayed biased retention/loss of 1 homeolog (Figure 2B, green lines). Except for keap1b, the biased retention/loss of other genes was shared by the two Carassius species and not asymmetrical to subgenome (five A-retained genes and six B-retained genes). The remaining four klhl genes with particularly uncertain cases were grouped in category IV (Figure 2B, gray lines).

We further investigated the association of the biased retention/loss of homeologs with the paralogs that were duplicates that generally resulted from the Ts3R and were usually named with the symbol of their mammalian orthologue followed by ‘a’/’b’ (e.g., klhl1a and klhl1b). A total of 42 klhl members with reliable homeologous and paralogous relationships were classified into four types (

Table 1. Four types of C. gibelio klhl genes based on the biased retention/loss of homeologs of paralogs.

Type	Gene Name	Gene Number	Biased Retention/Loss of Paralogs	Biased Retention/Loss of Homeologs
I	klhl20, klhl24, klhl29, klhl37, klhl39, klhl41	6	No	No
II	klhl10, keap1, klhl38, klhl40	4	No	Yes
III	klhl2, klhl4, klhl5, klhl6, klhl9, klhl11, klhl12, klhl13, klhl14, klhl15, klhl18, klhl21, klhl25, klhl26, klhl28, klhl30, klhl31, klhl32, klhl33, klhl34, klhl36, klhl43	22	Yes	No
VI	klhl3, klhl7, klhl13-like, klhl16, klhl22, klhl23.1, klhl23.2, klhl27, klhl35, klhl42	10	Yes	Yes

): Six klhl members showed no biased retention/loss of either paralog and homeolog (Type I); Four klhl members that did not lose any paralog had lost at least one homeolog (Type II); and Type III contained the largest number (22) of klhl members, of which one paralog had lost, but the remaining paralogs maintained all homeologs. The last type included 10 members, of which both paralogs and homeologs occurred biased retention/loss. Therefore, the paralogs that experienced biased retention/loss after the Ts3R tended to maintain homeologs after the allotetraploidy in Carassius.

2.3. Expression Atlas of klhl Genes in Carassius

To preliminarily infer their functions, we first investigated C. auratus klhl (Caklhl) gene expression across eight adult tissues and seven embryogenesis stages. Transcriptome results were verified as the consistent gene expression results of almost all 12 randomly selected klhl genes that were obtained by quantitative reverse transcription PCR (qPCR) analysis (Figure S1). Hierarchical clustering analysis showed that their expression patterns could be categorized into two distinct clusters (I and II) in adult tissues (Figure 3A): Cluster I included 47 genes that were expressed ubiquitously and averagely in most of the analyzed tissues; and Cluster II, containing the rest of the 49 genes, was subdivided into three subclusters (II1, II2, and II3) and mainly showed a tissue-preference expression pattern or low expression level. Genes in II1, such as klhl31-A, klhl31-B, klhl41b-A, and klhl41b-B, had much more expression in the muscle and/or heart than in other tissues. Genes in II2 showed a moderate expression level in one or several of the analyzed tissues. For example, klhl14-A and klhl14-B were specifically expressed in the brain, klhl32-B and klhl5-A were expressed in six tissues except the liver and muscle, and klhl26-B was mainly expressed in the ovary. The last subcluster contained eight genes that exhibited low expression level in almost all analyzed tissues. Similarly, based on the dynamic expression abundance during embryogenesis, Caklhl genes were also divided into two clusters (I and II) and each cluster was grouped into three subclusters. All 55 genes in cluster I showed abundant transcripts throughout embryogenesis, and most of them (e.g., klhl18s, klhl4s, and klhl13s) displayed the highest transcript abundance at the early stages (Figure 3C). Compared with cluster I, genes in cluster II were expressed at a lower level. Subcluster II1 genes gradually increased transcription abundance during embryogenesis. Consistent with the high expression level in the adult heart, a large amount of transcripts from both homeologs of klhl31, klhl40a, klhl41b, and klhl43 were detected from Prime-5 stage with the heart beating. Subcluster II2 included eight genes with a lower transcription level, and the transcripts of subcluster II3 genes were almost not detected (Figure 3C).

C. gibelio and C. auratus klhl orthologs showed very high sequence similarity: 68 out of 99 ortholog pairs appeared to have higher identities than 98% in the amino acid sequence, of which 20 ortholog pairs are exactly the same (Table S5). Also, consistent with their close evolutionary relationship, most of the Cgklhl genes had a very similar expression pattern as Caklhl orthologous genes (Figure 3). But there were still several genes displayed with an obvious difference between them. Compared to C. auratus adult tissues, for example, the Cgklhl11-A expression level decreased in all tissues, the Cgklhl16-B and Cgklhl37a-B expression levels were upregulated in the kidney, the Cgklhl42-A and Cgklhl38a-B expression levels increased in multiple tissues, and the expression of Cgklhl23.1-A was upregulated in the kidney, hypothalamus and pituitary (Figure 3B). Meanwhile, klhl42-A, klhl29a-A, and klhl11-A also showed differential expression patterns during embryogenesis (Figure 3D). Interestingly, except Cgklhl38a-B, all these differentially expressed genes showed very high identity, ranging from 98.8% to 100% (Table S5). Compared with the homologs in C. auratus, all C. giblio-specific duplicated copies (Cgklhl39-B1/B2 and Cgklhl8-1/2) showed similar expression patterns, and the total number of transcriptions were also comparable, suggesting their function may be conserved.

2.4. Varied Expression Bias of klhl Homeolog Pairs in Carassius

In polyploid, duplicated genes generally altered their expression patterns [17]. For combined analysis, in adult tissues, 68% of Caklhl and 47% of Cgklhl homeolog pairs showed balanced pattern (Figure 4A), with similar relative transcript abundance between homeologs A and B. The remaining klhl homeolog pairs belonged to homeolog-dominant patterns, with higher or lower expression from one homeolog relative to another. Nine klhl homeolog pairs showed consistent homeolog-dominant patterns (four with A-dominant and five with B-dominant) in Carassius (Figure S2). There were 11 homeolog pairs that expressed were bias expressed (six with A-dominant and five with B-dominant) exclusively in C. gibelio, whereas only three pairs (all with B-dominant) were C. auratus-specific (Figure 4B). However, during embryogenesis, a few more C. auratus klhl homeolog pairs showed bias expression in C. auratus than in C. gibelio (Figure 4A). Besides seven shared pairs (Figure S2), there were eight and six homeolog-dominant pairs exclusively in C. auratus and C. gibelio, respectively (Figure 4B). Among them, an interesting gene is klhl29a, which displayed homeolog B-dominant expression pattern in C. auratus but an A-dominant expression pattern in C. gibelio. In addition, these homeolog-dominant pairs usually expressed bias in multiple tissues or developmental stages rather than only one (Figure 4B and Figure S2).

We further investigated the expression bias of klhl genes in an individual sample. Consistent with the combined analysis, C. auratus and C. gibelio klhl genes both showed obvious homeolog B-dominant expression patterns in all adult tissues (Figure 4C, indicated by the medians). Except in the muscle and brain, Cgklhl genes displayed more homeolog B-dominant expression patterns in more tissues than Caklhl genes. In contrast, the expression levels of klhl homeologs were more balanced in C. gibelio during early embryogenesis (Figure 4D, E1–E4). From the 8-somite stage (E5), klhl genes displayed consistently homeolog B-dominant expression patterns in the two species. This may be related to the organogenesis and/or myogenesis in the late stages, during which the gene expression repertoire might be similar as in adult tissues. In addition, five homeolog pairs displayed a shift of expression dominance temporally during embryogenesis (Figure 4E). Except for klhl17a, four pairs showed an expression-bias shift both in C. auratus and C. gibelio, but their shifted patterns were not exactly the same. Four Caklhl genes shifted their expression bias between blastula (E2) to shield (E3) stages, while two genes (Caklhl41a and Cgklhl28) shifted between Prime-5 (E6) and Pec-fin (E7) stages.

2.5. klhl Expression in Response to Herpesvirus Infection in C. gibelio

Polyploids usually show enhanced tolerance to environmental or other stress [36]. To explore the functional divergence of klhl genes after autotriploidy in C. gibelio, we examined their expression in three C. gibelio gynogenetic clones infected by Carassius herpesvirus 2 (CaHV-2), one of the main pathogens in aquaculture of crucian carp [37,38]. The three clones diverged from a C. gibelio ancestor about 0.86 million years ago [19] but showed distinct herpesvirus resistance (clone A⁺: susceptible to CaHV-2; clone F: weakly resistant to CaHV-2; clone H: highly resistant to CaHV-2) [39]. Most of the klhl genes (70 out of 98) showed similar expression patterns among the three clones (Figure 5A). However, differentially expressed klhl genes (DEGs) in response to CaHV-2 infection showed an interesting pattern change among them. Clone H had 13 clone-specific DEGs, whereas clones A⁺ and F had only two clone-specific DEGs. Except for five downregulated DEGs (klhl4-A, klhl6-A, klhl9-B, klhl24a-B, and klhl39a-B) shared by all three clones, clone F shared eight and two DEGs with the same expression change with clone A⁺ and H, respectively, whereas clone A⁺ and H shared zero DEGs with similar expression change. In contrast, clone A⁺ and H had four DEGs that showed opposite expression changes after CaHV-2 infection (e.g., klhl5-A, klhl17b-B1, klhl24a-A, and klhl39b-B1). Correlation analysis indicated that clone A⁺ and H had the most distinct expression pattern for klhl genes in response to herpesvirus infection (r = −0.05), while clone F is in the middle of them (r = 0.77 for clone A⁺ and r = 0.13 for clone H) (Figure 5B).

Many differentially expressed klhl genes have been reported to be associated with cellular inflammation and immune regulation against viral infection [2,4]. For example, keap1a, a key sensor of oxidative stress and virus infection [40], was specifically upregulated in infected fish of clone H (Figure 5C), possibly related to the high resistance against CaHV-2; klhl12B, a suppressor to enterovirus-mediated translation [41], was downregulated in clone A⁺ and F (Figure 5D). Considering that Klhl proteins usually participate in the ubiquitination of substrate proteins via interacting with Cullin3 [4], other differentially expressed klhl genes may also be associated with the host’s pathological response to viral infection. Therefore, we speculate that the dynamic patterns of the differential expression of klhl genes would contribute to the different antivirus response among C. gibelio clones.

2.6. klhl Alleles in C. gibelio

To study variations among triploid homologs, we tried to identify Cgklhl alleles using a haplotype-resolved assembly of C. gibelio (see Materials and Methods) that contained 150 chromosomes and a read coverage analysis based on the assembly with 50 chromosomes (see details in methods). Of the total 98 Cgklhl genes, 90 genes possessed three alleles, whereas seven genes had two alleles and their average read depths (around ~36) were approximately 2/3 of the average depth for other Cgklhl genes (Table S6 and Figure 6A). Unexpectedly, Cgklhl18-B had only one allele, although its read depth was equal to the depth of genes with three alleles (Figure 6A): sequence comparing to the assembly with 150 chromosomes indicated that this gene had three other nonchromosomal homologous fragments that did not possess genes (Figure S3). We verified the alleles of all biallelic and monoallelic genes as well as 10 randomly selected triallelic genes, by PCR and Sanger sequencing using tissue cDNA mixture and genomic DNA templates. Therefore, allele loss of C. gibelio would be mainly caused by large fragment deletion.

Allelic sequence identities of the Cgklhl genes ranged from 88.12% to 100.00% (Table S6), and they were not significantly different between the genes belonging to subgenomes A and B (Figure S4). 81 out of 97 genes showed very high interallelic identities, higher than 99%. There were four genes (klhl4-A, klhl4-B, klhl12-A, and klhl29b-A) that showed interallelic identities lower than 90%, owing to small DNA fragment deletion in one of the alleles (Figure S5), but their structures of protein among alleles were not obviously different. Instead, the other three genes with high interallelic nucleotide identities had one allele that showed different protein domain structures (Figure 6B). One allele of klhl34-A lost most domains due to the small indels that caused a frame shift of the allele, whereas the single nucleotide polymorphisms resulted in missing one Kelch domain in one klhl42-A allele and two Kelch domains in two klhl23.2-A alleles.

2.7. Bias Allelic Expression of Cgklhl Genes

In general, the two alleles of diploid are expressed at similar levels, but some of them display unequal expression, referred to as allelic-specific expression (ASE) [42,43]. To understand how genetic interactions among alleles of amphitriploid influence gene expression, we analyzed the ASE of klhl genes in C. gibelio using STAR pipeline. The relative expression of three alleles determined a triad position in a ternary plot (Figure 7A, top left), which allows the analysis of allele expression bias. Accordingly, we defined three expression bias categories: a balanced category with similar transcript abundance among the three alleles (all alleles > 0.25); and suppressed or dominant categories, where one allele shows lower (one allele < 0.25, two alleles > 0.25) or higher (one allele > 0.5, two alleles < 0.25) transcript abundance than those from another two alleles. We first performed a global analysis that combined data across all seven adult tissues or eight developmental stages. In adult tissues, 10, 46, and 24 Cgklhl genes were classified into the balanced, suppressed, and dominant categories, respectively (Figure 7A); however, during embryogenesis, only two genes belonged to the balanced category and the dominant category included 45 genes (Figure 7A). The separated analysis for the individual sample also confirmed the allele expression bias, in which no gene belonged to the balanced category during the first five developmental stages (Figure 7B). A high proportion of unbalanced allele expression indicates that the alleles of most klhl genes have conspicuously diverged, especially during embryogenesis.

To explore the plasticity of the allele triads, we investigated their expression variation across the 15 transcriptome samples by calculating the mean distance from each triad position to the global average center (Figure 7C, left). In adult tissues, the mean distances of the 3 categories were around 0.097, 0.085, and 0.0081, respectively, and were not significantly different among them (all p-values > 0.1, two-sample t-test) (Figure 7C, right). Interestingly, the mean distances of all categories during embryogenesis (the medians were 0.31, 0.19, and 0.11, respectively) were significantly larger than those in adult tissues (all p-values < 0.05, two-sample t-test) (Figure 7C, right). These results indicate that the relative expression patterns of the three alleles are rather stable among adult tissues but dynamic during embryogenesis. Then, we hypothesized that the dynamic during embryogenesis might be associated with the allele dominance shift (ADS). To test this, we defined an allele as dominant when it accounts for more than 50% of the total transcript abundance in any of the samples, and examined the expression patterns of all dominant alleles. As shown in Figure 7D, no second allele was found to be dominant for all Cgklhl genes in all sampled adult tissues, whereas 29 out of 84 filtered triallelic genes had two dominant alleles for different developmental stages. Therefore, the ADS is indeed common during embryogenesis, suggesting that many alleles might function in a sequential manner. In addition, the ADS mainly took place between the 4-cell and blastula stages, corresponding to the maternal to zygotic transition, and between the 8-somite and 1-dpf stages, corresponding to the organogenesis.

3. Discussion

Whole genome sequencing and comparative genomic analyses have proved the impacts of WGD on genomic diversity, variability, and complexity [13,15,44,45]. The direct result of WGD is producing huge duplicates at the genome level. How the duplicates evolve their structure and expression for adapting to coexistence and for harmonizing is fascinating and needs extensive studies. It is critical for a new polyploid to alleviate conflicts of duplicates from genome mergers. Biased genomic changes may help mitigate the initial chaos and are essential for the subsequent rediploidization process [18,45]. The most likely evolutionary fate of redundant duplicated genes is loss of one duplicate, more than 70.0–80.0% of duplicated genes after Ts3R were lost in the first 60 million years [46]. Due to their relatively short evolutionary history, common carp (Cyprinus carpio) and goldfish did not experience extensive homeologous gene loss, but asymmetric subgenome evolution has been mentioned in both allotetraploids [47,48]. In this study, we found that 75% of klhl members have lost one paralog after Ts3R and 14 out of 55 klhl genes have lost one homeolog after 4R (Table S1 and Figure 1B). The klhl paralogs that experienced biased retention/loss are more likely to maintain homeologs. Dosage balance is likely a major driver for the patterns of duplicated gene retention and loss after polyploidization.

WGD provides flexibility to facilitate adaptive traits of species via tissue-specific expression or sub- or neofunctionalization of massively duplicated genes [15,49]. Differential homeolog expression, defined as homeolog bias, is a ubiquitous feature of plant polyploids [15,49]. For example, about 30.0% of hexaploid bread wheat (Triticum aestivum) homeolog triads (AABBDD subgenome) showed homeolog-dominant or homeolog-suppressed expression patterns [15,46]. However, besides expression bias, variant subgenome expression divergence processes have been observed in common carp and goldfish [50,51]. Versatile expression divergence is also found in the Carassius klhl gene family. Although expression dominance to homeolog B is in evidence, balanced homeolog pairs remain in the majority (Figure 4). Compared to adult tissues and late developmental stages, klhl homeolog pairs show more divergence in early developmental embryos, and some homeolog pairs even shift their expression bias during embryogenesis (Figure 4E). Interestingly, some homeolog pairs show different expression bias between C. auratus and C. gibelio although they share the same allotetraploidy event. These findings may be the result of the homeologous recombination and/or from the genome shock of the recent autotriploidy in C. gibelio. Polyploidy has been considered as an evolutionary and ecological force in stressful times or harsh environments, and polyploids often show enhanced tolerance to abiotic or biotic stress, including attacks by pathogens [36]. Relative to their ancestors, the increased tolerance of polyploids is thought to be attributable to several factors, such as genome buffering, gene expression altering, epigenetic remodeling, gene regulatory network rewiring, and sub- and/or neofunctionalization of duplicate genes [18,52]. The extra autotriploidy may endow C. gibelio with abilities for additional adaptive plasticity. Although most of the klhl genes had analogous expression patterns between C. gibelio and C. auratus (Figure 3), even including species-specific duplicated copies (Cgklhl39-B1/B2 and Cgklhl8-1/2) that also showed conserved expression patterns, there are still some genes that have undergone significant changes (Figure 3). Furthermore, this adaptive plasticity was also demonstrated by the divergent expression pattern among different C. gibelio gynogenetic clones infected with CaHV-2 (Figure 5). Although the three clones diverged from a C. gibelio ancestor about 0.86 million years ago [19], klhls expression changes in response to viral infection showed a clear correlation with different viral resistance among the three clones. Given the function of klhl gene in protein ubiquitination and the pathogenesis of diseases [4], the dynamic and variable expression divergences that improve functional plasticity and flexibility of duplicates may play important roles in evolutionary innovation for ecological adaptation, which awaits further investigation.

C. gibelio klhl genes are very similar in sequence and expression as C. auratus (Figure 3 and Table S5). In addition, no significant within-individual allelic diversity or decay were found in C. gibelio (Figure 6), although prolonged unisexual gynogenesis should naturally cause high homologous differentiation and accumulation of mutations [53]. The gene introgression/exchange with sympatric sexual species, especially with C. auratus, would help to shape the genome of C. gibelio. Recently, we found that changes in ploidy drive unisexual and sexual reproduction transition, and thereby facilitate genetic exchange from C. auratus to C. gibelio lineages [54]. This phenomenon has also been observed in many unisexual organisms, such as salamanders [54], Iberian-Roach [55], and water frogs [56] that can benefit from sampling the gene pools of the sexual relatives [57]. Moreover, homologous recombination during oogenesis may grant C. gibelio abilities to clean mutation and to keep genome stable [19]. Therefore, to achieve evolutionary longevity, C. gibelio might employ multiple strategies (from the interspecies and intraspecies) that not only refresh its genome but also increase the population genetic diversity.

Although homologous divergence is limited, the expression among alleles diverges somewhat from exclusive expression of one or two alleles to subtle expression differences among alleles (Figure 7). Interactions among alleles range from buffering effects when alleles are functionally redundant to dominance effects when variations in single/two alleles can lead to dominant phenotypes. This allelic expression imbalance is usually due to cis-acting genome variations because environmental or trans-acting elements commonly affect all alleles equally [42]. Besides transcription factor binding, chromatin accessibility mediated by epigenetic imprinting [58], random monoallelic expression, posttranscriptional regulation including nonsense-mediated decay, and alternative splicing [59,60] may also lead to allele-specific imbalance. Recently, another haplotype-resolved assembly of C. gibelio from the Olsa river close to Ostrava was reported [59,60]. Therefore, it will be valuable to study how haploids coordinate to evolve in this attractive fish and to compare these evolution features at the differential lineage background in the future.

In conclusion, we focused on the lineage specific expansion of the klhl gene family, which exerts a variety of critical functions by regulating ubiquitination of specific substrates, and elaborated the paralog/homeolog/homolog diversification, biased retention/loss, and differential expression across adult tissues and embryogenesis in Carassius. The current study will contribute to our understanding of fish klhl genes and provides a paradigm showing the evolutionary process of duplicated genes in complex genomes that experienced multiple rounds of polyploidization events. This understanding should increasingly inform ecological analyses of adaptation and permit an enhanced appreciation for genomic evolution of polyploidy-fueled diversification.

4. Materials and Methods

4.1. Identification and Characterization of klhl Family Genes

The genome assembly of C. auratus (GCA_019720715.2) and C. gibelio (GCA_019843895.2), as well as of other 11 species, including Cephalochordata amphioxus (Branchiostoma belcheri) (GCA_001625305.1), Sea lamprey (Petromyzon marinus) (GCA_010993605.1), elephant shark (Callorhinchus milii) (GCA_000165045.2), lobe-finned fish African coelacanth (Latimeria chalumnae) (GCA_000225785.1), chicken (Gallus gallu) (GCA_000002315.5), human (Homo sapiens) (GCA_000001405.28), Spotted gar (Lepisosteus oculatus) (GCA_000002035.4), zebrafish (Danio rerio) (GCA_000002035.4), fugu (Fugu rubripcm)(GCA_901000725.2), rainbow trout (Oncorhynchus mykiss) (GCA_901000725.2), and Atlantic salmon (Salmo salar) (GCA_905237065.2) were downloaded from available databases including Ensembl (http://asia.ensembl.org/index.html, accessed on 4 August 2021) and NCBI (https://www.ncbi.nlm.nih.gov/, accessed on 4 August 2021). Haplotype-resolved assembly of gibel carp was deposited in NGDC (https://ngdc.cncb.ac.cn/) under the accession number of PRJCA011199. First, the klhl members were identified through the annotation file for each genome. Second, Klhl protein sequences of humans and zebrafish were used for homology-based gene prediction using TblastN (blast-2.2.26), and the predicted genes were merged to the initial klhl genesets. Third, the hidden Markov model profiles of Klhl BTB (PF00651) and BACK (PF07707) domain (http://pfam.xfam.org/, accessed on 30 August 2021) were utilized as a query to verify and supplement the klhl genesets by using the HMMER 3.0 program with default parameters. The putative Cgklhl genes were named based on their reciprocal best homologs of humans and/or zebrafish. The complete cDNA sequences of 96 Caklhl genes (GenBank accession no. OL598607-OL598707) and 98 Cgklhl genes (GenBank accession no. OL677882-OL677984) were deposited in GenBank. The length of CDS/protein, isoelectric pointprotein (pI), and molecular weight (MW) of klhl genes were calculated by biopython module of python version 3.9.0.

4.2. Phylogenetic Analysis

Multiple alignments of the full length of Klhl amino acid sequences of C. auratus, C. gibelio, and D. rerio were achieved using ClustalW program of MEGA software (version 7.0) [61] with Neighbour-Joining (NJ) method and 1000 bootstraps. Subsequently, the phylogenetic tree was constructed and visualized using Figtree v1.4.3 (http://tree.bio.ed.ac.uk/software/Figtree/, accessed on 27 April 2021).

For allelic phylogenetic analysis, 90 triallelic klhl genes were selected to align with the orthologous Caklhl and Drklhl genes and construct phylogenetic trees with Neighbour-Joining (NJ) method and 1000 bootstraps.

4.3. Synteny Analysis

MCScanX was used to identify syntenic blocks between C. gibelio genome (GCA_019843895.2), C. auratus genome (GCA_019720715.2), and D. rerio genome (GCA_000002035.4) with the parameters of -a -e 1e-5 -u 1 -s 5. First, we used BLASTP to align proteins among the three genesets with the parameters “1 × 10⁻⁵”. The alignments were then subjected to MCScanX to determine syntenic blocks and visualized via TBtools (version 1.082) [62].

4.4. RNA-seq Samples and Gene Expression Analysis

RNA-seq data of eight adult tissues, embryos at seven developmental stages, and healthy/CaHV-2-infected head kidneys of three clones were downloaded from NCBI (PRJNA834570, PRJNA836313, PRJNA833750, PRJNA837728, and SRP096800). RNA-seq reads were mapped to the references using STAR [63] and the expression level was calculated using reads per kilobase per million mapped reads (RPKM). Pearson correlation coefficients were calculated in R 3.3.2.

The homeolog A/B expression ratio was calculated according to (log2 ((RPKM.A + 0.01)/(RPKM.B + 0.01))). The homeologous pairs were removed when the RPKM value of any gene was <5. The ratio of A to B expression more than 2 or less than 0.5 was defined as homeolog A- or B-dominant. For expression-bias shift analysis, A-dominant or B-dominant was defined when the ratio of A/B expression (Log2) was >0.1 or <−0.1.

Quantitative reverse transcription PCR (qPCR) was performed as previously described [64]. Heart and brain were sampled from adult C. gibelio. CaHV-2 infection and sample collection were performed as previously described [65] Specific primes for each gene were listed in Table S7. Actin was selected as the optimal reference gene. The relative gene expression levels were calculated with 2^−ΔΔCT method [66]. All the samples were analyzed in triplicate.

4.5. Polyploidy Evaluation and Allele Identification of Cgklhl Genes

BWA (Version 0.7.12-r1039) was used to map the Illumina reads of C. auratus and C. gibelio that we submitted to NCBI previously (accession numbers: OP253321-OP253605) to their genomes respectively and sorted to obtain the bam files by SAMtools (Version 1.4). The read depths for klhl genes were calculated using BamDeal (Version 0.25).

The CDS of Cgklhl genes were aligned to the haplotype-resolved assemblies (wang submitted) using blastn of TBtools (version 1.082). Combined with the ratio of average read depths of each gene to the average depth for other Cgklhl genes, to predict the number of alleles of Cgklhl genes, 1 corresponds to 3 alleles, 2/3 corresponds to 2 alleles, and 1/3 corresponds to 1 allele. We identified 285 klhl alleles for C. gibelio. Finally, PCR and 30 clones were sequenced to validate the alleles of randomly selected genes.

4.6. Allele-Specific Expression Analysis

Ninety triallelic klhl genes were selected for allele specific expression analysis. We defined a triad as expressed when at least 1 allele had the expression level >1 RPKM. After filtering, 80 and 84 triallelic klhl genes were left in the adult tissues and in the embryos, respectively, for further analyses. The relative expression of each allele across the triad was normalized by the total expression level of the gene. The triad was defined as ‘Dominant’ when the relative expression of one allele was >0.75, as ‘Suppressed’ when the relative expression of one allele was <0.25 and any of the other two alleles were <0.75, and as ‘Balanced’ when the relative expressions of all alleles were >0.25. In order to analyze the allele express shift, the allele expression values of the three alleles for all tissue or embryo samples were normalized firstly, and the allele with the highest expression level in a sample was defined as the dominant allele in this sample. By the end, the normalized expression levels of all dominant alleles were selected to drawn for all tissue or embryo samples. We calculated the Euclidean distance using the rdist function from R 3.3.2.

4.7. The Domains of Klhl Alleles

In order to predict protein conserved domains, the CgKlhl protein sequences were submitted to Pfam [67], NCBI-CDD [68] and SMART [69], and visualized by TBtools (version 1.082).

5. Conclusions

We explored the structure and expression evolutionary trajectories of klhl gene family in polyploid Carassius. Paralog/homeolog/homolog diversification, biased retain/loss, and versatile differential expression were elaborated in Carassius klhl gene family. The current study will contribute to our knowledge on fish klhl genes and provides a paradigm showing evolutionary process of duplicated genes in complex genomes that experienced multiple rounds of polyploidization events. This understanding should increasingly inform ecological analyses of adaptation and permit an enhanced appreciation for genomic evolution of polyploidy-fueled diversification.