Convergent copy number increase of genes associated with freshwater colonization in fishes

Abstract

Copy number variation (CNV) can cause phenotypic changes. However, in contrast to amino acid substitutions and cis-regulatory changes, little is known about the functional categories of genes in which CNV is important for adaptation to novel environments. It is also unclear whether the same genes repeatedly change the copy numbers for adapting to similar environments. Here, we investigate CNV associated with freshwater colonization in fishes, which was observed multiple times across different lineages. Using 48 ray-finned fishes across diverse orders, we identified 23 genes whose copy number increases were associated with freshwater colonization. These genes showed enrichment for peptide receptor activity, hexosyltransferase activity and unsaturated fatty acid metabolism. We further revealed that three of the genes showed copy number increases in freshwater populations compared to marine ancestral populations of the stickleback genus Gasterosteus. These results indicate that copy number increases of genes involved in fatty acid metabolism (FADS2), immune function (PSMB8a) and thyroid hormone metabolism (UGT2) may be important for freshwater colonization at both the inter-order macroevolutionary scale and at the intra-genus microevolutionary scale. Further analysis across diverse taxa will help to understand the role of CNV in the adaptation to novel environments.

This article is part of the theme issue ‘Genetic basis of adaptation and speciation: from loci to causative mutations’.

1. Introduction

Parallel and convergent evolution of phenotypic traits in independent lineages that inhabit similar environments is prevalent in nature, indicating the role of natural selection and the repeatability of evolution at the phenotypic level [1–3]. Recent genetic studies have further revealed that mutations in the same genes often occur when organisms are adapted to similar environments, which is called genetic parallelism or genetic convergence [4,5]. There are many striking examples of amino acid substitutions and cis-regulatory changes in the same genes for adapting to similar environments [4–9]. Copy number variation (CNV) is another type of mutation that can lead to phenotypic changes through a variety of mechanisms such as changes in gene expression levels, reorganization of chromatin structure, and subsequent mutations at the cis-regulatory regions and amino acid sequences [10–16]. Several studies have shown that the same genes show repeated increases of copy numbers in organisms treated with antibiotics and pesticides [11] and mammals adapted to starch-rich diets during domestication by the duplication of amylase genes [17,18]. By contrast, there are only a few reports of CNV underlying adaptation in natural systems except for a few cases, such as the adaptation to freshwater habitats by the duplication of fatty acid desaturase genes in fishes [19] and the evolution of floral pigmentation by the duplication of anthocyanin-regulating transcription factors in plants [20]. Therefore, we do not know how prevalent parallel and convergent CNV is in nature.

Fishes provide us an excellent opportunity to test whether any genes exhibit parallel and convergent copy number changes during transitions from marine to freshwater environments, because freshwater colonization occurs repeatedly in multiple independent lineages [21]. Recently, we found that the copy number increases of the fatty acid desaturase gene FADS2, a gene that encodes an enzyme for the biosynthesis of docosahexaenoic acid (DHA), repeatedly underlie freshwater colonization in fishes [19]. Freshwater ecosystems are generally poor in DHA, an omega-3 long-chain polyunsaturated fatty acid (LC-PUFA), which plays crucial roles in growth, survival and reproduction in animals [22]. An increase in the copy number of FADS2 can help fishes overcome DHA deficiency in freshwater habitats. Importantly, FADS2 amplification associated with freshwater colonization was observed not only at the macroevolutionary scale examining 48 species across diverse orders of ray-finned fishes, but also at the microevolutionary scale examining inter-population variations within species. The three-spined stickleback, Gasterosteus aculeatus, is primarily a marine fish species, but repeatedly colonized freshwater habitats on multiple continents [23]. Freshwater populations of the three-spined stickleback were found to have higher FADS2 copy numbers than marine populations across multiple geographical regions. Since marine and freshwater habitats differ not only in fatty acid availability, but also in many other environmental factors, such as salinity, water flow, pathogens and predators, genes other than FADS2 may also show increased copy numbers in freshwater fishes compared to marine fishes.

In the present study, we first tested whether any genes other than FADS2 showed convergent copy number changes in the freshwater fishes than in the marine fishes across orders using whole-genome sequences of 48 ray-finned fishes. Next, we tested whether any of these genes also showed copy number increases associated with freshwater colonization within the stickleback genus Gasterosteus in particular. Previous studies have listed genes showing copy number increases associated with freshwater colonization in the three-spined stickleback [24,25]. However, this work used only North American and European freshwater stickleback populations; thus, no Asian freshwater populations were included in the analyses. Because the genetic basis of freshwater adaptation often differs across geographical regions [26–32], we investigated the CNV among Japanese stickleback populations in the present study and compared our results with previous studies on the North American and European stickleback populations [24,25]. Finally, for the several genes whose copy number increases were associated with freshwater colonization, we investigated whether gene amplification occurred by tandem duplication in a freshwater stickleback population.

2. Methods

(a) . Association between copy number and habitat in the ray-finned fishes

To identify genes that show copy number increases associated with freshwater colonization, we examined the protein sequence data of 48 ray-finned fishes, which were available on Ensembl (https://asia.ensembl.org/index.html) or RefSeq (https://www.ncbi.nlm.nih.gov/refseq/). When the gene ID has several splicing variants, we used a protein sequence from the longest splice variant, ensuring that the dataset contained one peptide sequence for each gene ID. The species were first classified into ‘freshwater species’ that have freshwater populations (34 species) and ‘non-freshwater species' that lack any freshwater populations (14 species) based on Eschmeyer's catalog of fishes [33]. Next, we determined the orthologous relationships of all genes among these species using the SonicParanoid v. 1.2.6 [34]. SonicParanoid was executed with the ‘default’ mode (i.e. the sensitivity parameter for sequence search was set to 4). To identify genes whose copy number increases are strongly associated with freshwater colonization, we selected genes that met the following three criteria: (i) the average copy number of freshwater species is more than twice that of non-freshwater species, (ii) more than 90% of freshwater species have at least one copy of the gene and (iii) more than 70% of non-freshwater species have less than two copies of the gene. Orthologue gene IDs were assigned in the order of copy number ratio between freshwater and non-freshwater species (table 1; electronic supplementary material, table S1; figure 1).

Table 1.

Candidate genes that show increased copy numbers associated with freshwater colonization after phylogenetic correction.

orthologuos gene ID	gene name	protein name	protein function	references for protein function	mean copy number in freshwater species	mean copy number in non-freshwater species	freshwater/marine CNV ratio	pMCMC	pMCMC (with WGD)
5	LTBP1/4	latent transforming growth factor beta binding protein 1	pulmonary, gastrointestinal, urinary, musculoskeletal, craniofacial and dermal development	[35]	1.117647059	0.285714286	3.911764706	0.0136	0.0124
7	TMEM150A		localization of phosphatidylinositol 4-kinase (PI4K) to the plasma membrane	[36]	1.352941176	0.357142857	3.788235294	0.0092	0.0064
9	YWHAQ	14-3-3 protein theta	signalling pathways, including metabolism, cell division, stress responses, protein trafficking, and immune responses, insulin sensitivity	[37,38]	1.647058824	0.5	3.294117647	0.032	0.0376
11	NFIA	nuclear factor IA	brain maturation, spinal cord development	[39,40]	1.088235294	0.357142857	3.047058824	0.0128	0.0116
14	PTH2RA	parathyroid hormone 2 receptor A	calcium and bone homeostasis, corticosterone release, anxiety state, fear response, thermoregulation, prolactin, postnatal pup development	[41,42]	1.647058824	0.571428571	2.882352941	0.0352	0.0272
21	HSF5	heat shock transcription factor 5	spermatogenesis	[43]	1.735294118	0.714285714	2.429411765	0.0372	0.0348
22	FUT9A	fucosyltransferase 9A	tissue development, angiogenesis, fertilization,cell adhesion, inflammation and tumour metastasis	[44]	2.235294118	0.928571429	2.407239819	0.0048	0.0052
23	DTX3L	deltex E3 ubiquitin ligase 3L	DNA damage repair, immune system	[45,46]	2.058823529	0.857142857	2.401960784	0.0368	0.0392
24	CART1	cocaine-and amphetamine-regulated transcript protein 1	feeding behaviour	[47]	1.029411765	0.428571429	2.401960784	0.034	0.0396
25	CXCR3	C-X-C motif chemokine receptor 3	immune system	[48]	2.147058824	0.928571429	2.312217195	0.0444	0.0368
27	PTGR1	prostaglandin reductase 1-like	inflammation, fatty acid metabolism	[49–51]	2.411764706	1.071428571	2.250980392	0.004	0.0072
28	UGT2	UDP-glucuronosyltransferase 2	drug metabolism, thyroid homeostasis, glucuronidation	[52–55]	3.852941176	1.714285714	2.24754902	0.0156	0.0124
29	CAPN2	calpain-2 catalytic subunit	early embryonic development	[56,57]	1.441176471	0.642857143	2.241830065	0.0236	0.0268
33	FADS2	fatty acid desaturase 2	DHA biosynthesis	[19]	2.205882353	1	2.205882353	0.0032	0.0052
36	PATJ	PALS1-associated TJ protein	tight junction formation, cell polarization	[58,59]	1.088235294	0.5	2.176470588	0.0056	0.0072
38	NAV2A	neuron navigator 2A	neurogenesis	[60]	1.529411765	0.714285714	2.141176471	0.0472	0.082
40	si:dkey-22o22.2	si:dkey-22o22.2		[61]	1.058823529	0.5	2.117647059	0.0284	0.0328
41	CXorf40A	CXorf40A	inflammatory responses	[62]	1.205882353	0.571428571	2.110294118	0.0184	0.0212
43	CD4	T-cell surface glycoprotein CD4	immune system	[63,64]	1.794117647	0.857142857	2.093137255	0.0276	0.034
49	PSMB8a	proteasome 20S subunit beta 8a	immune system	[65]	2.029411765	1	2.029411765	0.0488	0.0684
54	LAMC3	laminin, gamma 3	neurogenesis	[66]	1.147058824	0.571428571	2.007352941	0.026	0.0144
55	DOC2B	double C2 domain beta	neurotransmission, insulin secretion, insulin sensitivity	[67,68]	1.147058824	0.571428571	2.007352941	0.0348	0.0328
56	USP47	ubiquitin carboxyl-terminal hydrolase 47-like	synapse development, behaviour, immune system	[69–72]	2.294117647	1.142857143	2.007352941	0.0412	0.0548

Figure 1.

Genes showing increased copy numbers in freshwater ray-finned fishes. The left panel indicates the phylogenetic tree of ray-finned fishes, which is primarily based on Betancur-R. [73] (see Methods). The light blue and pink letters indicate species with freshwater populations and those that lack freshwater populations, respectively. Circles at nodes indicate bootstrap support values in Betancur-R. [73]. The right panel shows a heat map of the gene copy numbers. The light and deep blue squares indicate lower and higher copy numbers, respectively. The orthologous gene IDs at the bottom correspond to the orthologous gene IDs in table 1 and electronic supplementary material, table S1. The orthologue gene IDs were assigned in the order of copy number ratio between freshwater and non-freshwater species. The asterisks indicate significant increases in copy numbers in freshwater species compared to non-freshwater species after phylogenetic correction (pMCMC < 0.05).

Although the main focus of the present study was to identify candidate freshwater-adaptive genes that increased their copy numbers, we also searched for genes with significantly lower copy numbers in freshwater species than in non-freshwater species. We selected genes that met the following three criteria: (i) the average copy number of non-freshwater species is more than twice that of freshwater species, (ii) more than 90% of non-freshwater species have at least one copy of the gene and (iii) more than 70% of freshwater species have less than two copies of the gene (electronic supplementary material, table S2).

To exclude the possibility that the observed associations between gene copy number and habitat simply reflects their phylogenetic relationships [74], we conducted Bayesian inference for a generalized linear mixed model (GLMM). This accounted for phylogeny as a covariate using the MCMCglmm R package [75] with the published fish phylogenetic tree [73], as described previously [19]. The estimated copy number of each gene was used as a response variable, while the habitat type (freshwater versus non-freshwater species; see above) was used as a predictor.

To investigate what kinds of genes showed CNV, we performed gene ontology (GO) analysis. For the GO term enrichment test of genes that increased in freshwater species, we used the stickleback orthologues, because the sticklebacks contain freshwater populations. By contrast, for the GO term enrichment test of genes that increased in non-freshwater species, we used the spiny choromis (Acanthochromis polyacanthus) orthologues, because the spiny choromis is entirely marine. For genes that had multiple copies with different gene IDs, we randomly selected one gene ID to be conservative. The GO enrichment analysis was performed against all annotated stickleback genes using g: GOSt in g: Profiler (version e104_eg51_p15_3922dba) [76]. We listed GO terms that were significantly enriched (Benjamini–Hochberg FDR p < 0.05) and found in at least two genes in the query.

(b) . Analysis of copy number variation between marine and freshwater sticklebacks

We explored whether the genes whose copy number increases were found to be associated with freshwater colonization at the macroevolutionary scale (see above) showed similar copy number increases in freshwater populations of the three-spined stickleback compared to marine populations. Previous studies have reported a list of genes that show copy number increases associated with freshwater colonization of the three-spined stickleback [24,25]. In these studies, whole-genome sequences of 11 freshwater and 10 marine stickleback populations [77] were analysed [24,25]. Hirase et al. [25] reported 19 genes, but Lowe et al. [24], having developed a novel method for detecting CNV, reported 138 genes, including all 19 genes identified in Hirase et al. [25]. Thus, we investigated whether any of these 138 genes overlapped with genes for which we had found increased copy numbers in freshwater fishes at the macroevolutionary scale.

Because previous studies on sticklebacks did not include freshwater populations from Asia [24,25], we investigated whether Japanese freshwater populations shared copy number increases with the North American and European freshwater populations. Furthermore, we analysed multiple individuals per population, which extends our analytical scope relative to previous studies that sequenced only one individual per population [24,25]. We investigated the copy number of candidate genes using our previously reported whole-genome sequences (WGS) of 75 individuals [19]: two marine populations (five females and five males of Japan Sea stickleback from Akkeshi, six females and four males of Pacific Ocean marine population of the three-spined stickleback from Akkeshi) and four freshwater populations (five females and three males from Nishitappu, six females and two males from Chimikeppu, three females and eight males from Gifu, three females and six males from Ono). Genomic DNA was isolated using the Qiagen DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA). The libraries were constructed using the NEBNext Ultra DNA Library Prep Kit (NEB, Ipswitch, MA, USA) and run in a 150 bp-paired end mode of HiSeqX. Sequence data are available from DDBJ (DRA007515 and DRA001136). The obtained sequence reads were trimmed and mapped to the BROADS1 reference sequence and cDNA coding sequence of each candidate gene, which spanned from the first ATG to the stop codon without introns, using the CLC Genomics Workbench 10.1.1 as described previously [19].

To test whether similar copy number increases occurred in the Japanese freshwater populations compared to the North American and European freshwater populations, we focused on five genes, FADS2 (ENSGACG00000005442), PSMB8a (ENSGACG00000000146), UGT2 (ENSGACG00000013874), PTGR1 (ENSGACG00000019130) and USP47 (ENSGACG00000019203), which possess two or more copies in the BROADS1 stickleback reference genome, which is the sequence of an Alaskan freshwater population belonging to the same Pacific stickleback lineage [28]. Since each candidate gene has multiple cDNA sequences, we conducted a phylogenetic analysis of cDNA sequences using the CLC Genomics Workbench and used the cDNA sequence located at the basal position of the phylogenetic tree for mapping. To calculate the copy number, the coverage of the cDNA was divided by the genome-wide coverage for each individual. All statistical analyses were conducted using the R v. 3.6.2 [78]. To test whether marine and freshwater populations possess different copy numbers for each gene, we used GLMM with a gamma distribution with the glmer function of the lme4 package in R [79], that account for the relative copy number of each gene per individual as the dependent variable, the ecotype (marine or freshwater) and the sex as the independent variables, and the population as the random effect.

(c) . Local genomic similarity analyses around candidate freshwater-adaptive genes in a freshwater stickleback population

To investigate whether duplicated copies were tandemly arrayed, we analysed raw reads of previously generated PacBio long-read sequencing data of a Gifu freshwater population (available at DRA007518) [19]. We conducted BLAST searches to find sequence reads containing FADS2, PSMB8a and UGT2, using the CLC Genomics Workbench. The cDNA sequences used in our search query were as follows: FADS2, ENSGACG00000005442; PSMB8a, ENSGACG00000000146; UGT2, ENSGACG00000013874. The following parameters were used: number of threads = 1, expect = 10, word size = 11, match = 2, mismatch = −2, and gap cost = Existence 5, Extension 2. Local pairwise alignment was conducted using the YASS [80]. Repetitive sequences were identified by searching against a database of known repetitive sequences using the GIRI Repbase in CENSOR software (http://www.girinst.org/censor/index.php) [81].

3. Results

(a) . Genes that show copy number differences between freshwater and non-freshwater fishes

We first screened for genes that showed increased copy numbers in freshwater species compared to non-freshwater species. We identified 58 candidate genes that showed increased copy numbers associated with freshwater colonization (figure 1; electronic supplementary material, table S1). GO analysis indicated enrichment of genes related to G protein-coupled receptor activity, immune receptor activity, cytokine receptor activity, chemotaxis and unsaturated fatty acid metabolism (electronic supplementary material, table S3). Of the 58 genes, 23 genes showed statistical significance even after phylogenetic correction (pMCMC < 0.05; table 1). The FADS2 gene, which we previously found to increase in copy number in freshwater fishes [19], was included in this list. GO analysis of these 23 genes showed enrichment of genes involved in peptide receptor activity, hexosyltransferase activity and unsaturated fatty acid metabolism (electronic supplementary material, table S4).

We identified 18 genes that showed higher copy number in non-freshwater species compared to freshwater species (electronic supplementary material, table S2). GO analysis of these 18 genes showed enrichment of genes involved in carboxylic ester hydrolase activity, catalytic activity, endopeptidase activity and methyltranferase activity (electronic supplementary material, table S5). Of these 18 genes, 12 genes exhibited statistical significance even after phylogenetic correction (pMCMC < 0.05). GO analysis of these 12 genes revealed enrichment of genes involved in carboxylic ester hydrolase activity (electronic supplementary material, table S6).

(b) . Copy number increase in freshwater sticklebacks

Among the 58 candidate freshwater-adaptive genes described above (electronic supplementary material, table S1), three genes, FADS2, GVINP1 and CXCR1, overlapped with those that were previously reported to show higher copy numbers in North American and European three-spined stickleback freshwater populations compared to marine populations (electronic supplementary material, table S1) [24]. Only the FADS2 gene was included in the 23 genes showing a significant difference after phylogenetic correction.

To test whether the copy number increases occurred in the Japanese freshwater populations, we focused on five genes, FADS2, PSMB8a, UGT2, PTGR1 and USP4, which possess two or more copies in the Alaskan freshwater population belonging to the same Pacific stickleback lineage [28]. In the Japanese freshwater three-spined stickleback populations, FADS2, PSMB8a and UGT2 showed significantly higher copy numbers compared to marine populations (GLMM: p = 0.026, p = 0.012 and p = 0.0023, respectively; figure 2; electronic supplementary material, table S7). USP47 and PTGR1 did not show significant differences between habitats (GLMM: p = 0.37 and p = 0.35, respectively), although there are inter-population variations such that Pacific Ocean marine and Gifu freshwater populations possessed significantly lower copy numbers of USP47 and PTGR1, respectively, than other populations (GLM: p = 0.022 and p = 0.0021, respectively).

Figure 2.

Estimated copy number of the FADS2, PSMB8a, UGT2, USP47 and PTGR1 genes in marine and freshwater sticklebacks. The pink and light blue boxes indicate marine and freshwater sticklebacks, respectively. Each dot indicates a single individual. Since at least one copy of FADS2 gene is located on the X chromosome, the relative copy numbers of FADS2 are shown separately for each sex.

To investigate how the copy number increases have occurred in the freshwater stickleback populations, we analysed the long-read sequencing data of a Japanese freshwater stickleback (Gifu population). We had already found that the FADS2 gene is tandemly duplicated with many transposons in Japanese and North American populations [19]. We also found tandem duplication of UGT2 genes (figure 3). Local genomic similarity analyses revealed that at least six copies of the UGT2 genes were tandemly duplicated. The similarity analyses also detected accumulation of transposons around the tandemly duplicated copies (figure 3; electronic supplementary material, table S8).

Figure 3.

Local pairwise alignment of genomic regions around the UGT2 gene against themselves. The three longest PacBio sequence fragments containing the UGT2 gene are shown (a–c). The green dots represent forward alignments, and the red dots represent reverse alignments; however, the red dots are so rare at these regions that they cannot be seen. The blue arrows indicate the UGT2 genes. The upper bar indicates the positions of repetitive elements identified in Repbase by CENSOR. The red-purple, yellowish-brown, blue and purple bars indicate DNA transposons, long terminal repeat (LTR) transposons, non-LTR transposons and interspersed repeats, respectively. The list of transposons is presented in electronic supplementary material, table S8.

4. Discussion

(a) . Gene amplification associated with freshwater colonization across orders of ray-finned fishes

Our screening for genes whose copy number amplification was associated with freshwater colonization across ray-finned fish orders identified 58 candidate genes. Twenty-three genes remained significant even after phylogenetic correction. Importantly, FADS2, a previously identified gene that shows a copy number increase in freshwater fishes [19], was included in these 23 candidate genes, suggesting that our screening worked well. By contrast, 18 genes showed higher copy number in non-freshwater species, and 12 genes remained significant after phylogenetic correction. Therefore, the number of genes that increased the copy numbers was almost twice as much as that of those that decreased in freshwater species.

GO analysis showed that genes involved in unsaturated fatty acid metabolism and immune functions were enriched in genes that showed copy number increases in freshwater fishes. This is consistent with the hypothesis that increased expression is often beneficial for proteins involved in interaction with the environment, such as stress response and metabolism [14]. Because there is variation in the availability of omega-3 and omega-6 LC-PUFAs between marine and freshwater ecosystems [82], increased copy numbers of PTGR1, which is involved in arachidonic acid metabolism, may be advantageous in freshwater species, which is also the case for the FADS2 gene. Furthermore, freshwater fish species may be exposed to strong selective pressures due to diverse and abundant parasites and pathogens relative to marine fishes [83,84]. Therefore, gene duplication of immune system-related genes may serve as a reservoir from which new genes constantly arise to protect against diverse pathogens [85]. Gene copy number increases concerned with fatty acid metabolism and immune functions may facilitate freshwater colonization and adaptation in ray-finned fishes.

By contrast, genes involved in carboxylic ester hydrolase activity were enriched in genes that showed higher copy number in non-freshwater species. The carboxylic ester hydrolase, also called bile salt-dependent lipase, can hydrolyse wax esters [86,87]. Wax esters are abundant in marine zooplankton, especially the dominating copepods, but generally more resistant to hydrolysis by pancreatic lipase, which can hydrolyse triacylglycerol substrates [87]. Since marine fishes feed mainly on zooplankton including calanoid copepods enriched with wax esters, copy number increases of genes related to carboxylic ester hydrolase activity may help them to use wax esters in marine environments.

(b) . Gene amplification associated with freshwater colonization within the genus Gasterosteus

Three genes (FADS2, GVINP1 and CXCR1) showed copy number increases in North American and European freshwater stickleback populations, whereas FADS2 and two other genes (PSMB8 and UGT2) showed copy number increases in Japanese freshwater stickleback populations. Among these, three genes, PSMB8, GVINP1 and CXCR1, were related to immune functions. PSMB8 encodes a catalytic subunit of the immunoproteasome responsible for generating peptides presented by major histocompatibility complex (MHC) class I molecules [65]. PSMB8 knockout mice show reduced expression of MHC class I molecules on cell surfaces [86]. GVINP1 encodes an interferon-induced very large GTPase 1 [87]. GVINP is known to play a role in immune reactions against pathogens [88]. In Atlantic salmon, the GVINP1 gene is located in a QTL region that explains over 20% of the genetic variance in resistance to amoebic gill disease and is differentially expressed between resistant and susceptible individuals [89]. The CXCR1 gene encodes a C-X-C motif chemokine receptor 1, which is a G protein-coupled receptor for the CLC chemokine interleukin-8 (IL-8), a major mediator of immune and inflammatory responses [90]. The CXCR1 gene plays a crucial role in the IL-8 signal transduction pathway of neutrophils [91–93]. A genome-wide association study revealed that the CXCR1 gene is located at a genomic region explaining the number of piglets born alive, which is potentially affected by CXCR1 immune function [94]. Therefore, high copy numbers of PSMB8a, GVINP1 and CXCR1 genes may contribute to adaptation to the diverse pathogens in freshwater environments.

UGT2 encodes UDP-glucuronosyltransferase 2, which metabolizes and inactivates triiodothyronine (T3), an active form of thyroid hormone found in the liver [55,95,96]. Thyroid hormones play key roles in the regulation of many physiological and behavioural processes, such as metabolism, ion homeostasis, basal activity, growth and development. Importantly, previous studies demonstrated that freshwater sticklebacks possess lower plasma concentrations of T3 and T4 than marine sticklebacks at both juvenile [97] and adult stages [30]. Lower levels of thyroid hormone are correlated with a lower metabolic rate which is likely adaptive in oxygen- and nutrient-limited freshwater environments [30]. Different expression levels of thyroid stimulating hormone ß2 (TSHß2), which may stimulate the synthesis and secretion of thyroid hormones, and iodothyronine deiodinase genes, DIO2 and DIO3, which convert T4 to T3 and T3 to T4, respectively, may underlie this divergence in thyroid hormone levels [30,98]. The higher copy numbers of UGT2 may also contribute to the lower thyroid hormone levels in freshwater sticklebacks. Even in teleosts other than sticklebacks, there are several reports on inter-population and geographical variation in thyroid hormones at adult stages [98]. Thyroid hormones are also deposited in egg of teleosts [99,100]. The eggs of freshwater fishes also contain lower concentrations of T3 than those of non-freshwater fishes [101]. Because freshwater fish larvae have lower metabolic requirements [102], the higher copy numbers of UGT2 in freshwater fishes may be responsible for the lower yolk thyroid hormone levels and lower metabolic activity at the larval stage.

The long-read genome sequencing data revealed tandem duplication of the UGT2 genes with surrounding transposons. It is similar to the FADS2 gene in freshwater sticklebacks, which is tandemly duplicated with many transposons [19]. Because transposons can induce tandem sequence duplications in plants [103], transposons near the UGT2 gene might have facilitated these tandem duplications. Because we did not conduct polishing of PacBio sequences by Illumina short reads, the currently available sequences are not of high enough quality for pinpointing the exact breakpoints of the UGT2 duplication. Precise identification of the boundaries of the duplicated regions with the UGT2 gene clusters may help to understand how adaptive tandem duplication occurs.

(c) . Genetic parallelism versus genetic non-parallelism with regard to copy number variation

Three genes (FADS2, PSMB8 and UGT2) showed a convergent increase in copy number during freshwater colonization, both at the macroevolutionary scale across multiple orders and at the within-genus microevolutionary scale. However, there were many genes that did not overlap between them. Furthermore, even among the stickleback lineages, different geographical populations showed different patterns. For example, PSMB8 and UGT2 showed increased copy numbers only in the Japanese freshwater stickleback populations, not in the North American and European freshwater stickleback populations. These results suggest that the genetic basis for freshwater adaption may differ between lineages and geographic regions.

Consistent with our findings, previous studies have reported that the genetic basis of freshwater adaptation often differs across geographical regions in sticklebacks [26–32]. Furthermore, a meta-analysis indicated that the probability of using the same genes for parallel and convergent phenotypic evolution declines with genetic divergence [104]. Therefore, it is not surprising that different fish lineages use different CNV to adapt to similar environments. Further research regarding CNV underlying repeated adaptation in natural populations will shed light on the generality and lineage specificity in recurrent gene copy number increases.

5. Conclusions

We have identified several genes that show convergent increases in copy numbers during freshwater colonization by analysing macroevolutionary patterns of gene copy numbers across orders of ray-finned fishes. Among freshwater fishes, candidate genes showing increased copy numbers are involved in fatty acid metabolism and immune function. Some of the CNV were also observed even within a genus of Gasterosteus. Currently, little is known about how the identified CNV impact phenotypic traits that help fishes colonize freshwater environments. Further studies on the functional roles of increased copy number of these candidate genes will help to understand the genetic mechanisms of freshwater colonization that occurred repeatedly in fishes.

Ethics

All animal experiments were approved by the institutional animal care and use committee of the National Institute of Genetics (R2-18). Before dissection, all fish were euthanized with 500 mg l⁻¹ MS-222.

Data accessibility

The whole-genome sequence data are available from DDBJ DRA007515, DRA001136 and DRA007518. The datasets supporting this article have been uploaded as part of the electronic supplementary material [105]. Electronic supplementary material, table S1: candidate genes that show increased copy numbers associated with freshwater colonization. Electronic supplementary material, table S2: candidate genes that show increased copy numbers in non-freshwater fish species. Electronic supplementary material, table S3: GO enrichment analysis for 58 candidate genes that show increased copy numbers associated with freshwater colonization. Electronic supplementary material, table S4: GO enrichment analysis for 23 candidate genes that remained significant even after phylogenetic correction. Electronic supplementary material, table S5: GO enrichment analysis for 18 candidate genes that show increased copy numbers in non-freshwater species. Electronic supplementary material, table S6: GO enrichment analysis for 12 candidate genes that show increased copy numbers in non-freshwater species and remained significant even after phylogenetic correction. Electronic supplementary material, table S7: candidate genes that show convergent increases of copy number in Japanese freshwater sticklebacks. Electronic supplementary material, table S8: transposons near the UGT2 gene in a freshwater stickleback. Electronic supplementary material, figure S1: High-resolution image of figure 3.

Authors' contributions

A.I.: conceptualization, data curation, formal analysis, funding acquisition, investigation, validation, visualization, writing—original draft; S.Y.: data curation, formal analysis, visualization, writing—review and editing; W.I.: funding acquisition, writing—review and editing; J.K.: conceptualization, funding acquisition, project administration, supervision, writing—original draft.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This project was supported by JSPS KAKENHI grant nos. 19H03277, 19KK0187 and 20H04873, the Inamori Research Grants to A.I., JST CREST JPMJCR19S2 to W.I. and JSPS KAKENHI grant no. 19H01003 to J.K.