Chromosome‐Level Genome Assembly of the Loach Goby Rhyacichthys aspro Offers Insights Into Gobioidei Evolution

ABSTRACT

The percomorph fish clade Gobioidei is a suborder that comprises over 2200 species distributed in nearly all aquatic habitats. To understand the genetics underlying their species diversification, we sequenced and annotated the genome of the loach goby, Rhyacichthys aspro , an early‐diverging group, and compared it with nine additional Gobioidei species. Within Gobioidei, the loach goby possesses the smallest genome at 594 Mb, and a rise in species diversity from early‐diverging to more recently diverged lineages is mirrored by enlarged genomes and a higher presence of transposable elements (TEs), particularly DNA transposons. These DNA transposons are enriched in genic and regulatory regions and their copy number increase is strongly correlated with substitution rate, suggesting that DNA repair after transposon excision/insertion leads to nearby mutations. Consequently, the proliferation of DNA transposons might be the crucial driver of Gobioidei diversification and adaptability. The loach goby genome also points to mechanisms of ecological adaptation. It contains relatively few genes for lateral line development but an overrepresentation of synaptic function genes, with genes putatively under selection linked to synapse organisation and calcium signalling, implicating a sensory system distinct from other Gobioidei species. We also see an overabundance of genes involved in neurocranium development and renal function, adaptations likely connected to its flat morphology suited for strong currents and an amphidromous life cycle. Comparative analyses with hill‐stream loaches and the European eel reveal convergent adaptations in body shape and saltwater balance. These findings shed new light on the loach goby's survival mechanisms and the broader evolutionary trends within Gobioidei.

1. Introduction

The order Gobiiformes represents a remarkable array of diversity within vertebrates, encompassing over 2200 species across 321 genera, ranking it among the largest vertebrate orders (Kuang et al. 2018; Nelson et al. 2016). These fish occupy all aquatic environments on earth (Brandl et al. 2018, 2019). Most species are marine, while around 10% are found in freshwater habitats (Sidlauskas and Chakrabarty 2010). Characterised by their extraordinary variety of forms, ecological roles, and behaviours, Gobiiformes have become prominent model organisms in comparative research, shedding light on the evolutionary mechanisms driving species diversification (Agorreta et al. 2013; de Brito et al. 2022). The order Gobiiformes includes nine families. Eight of them, Rhyacichthyidae, Odontobutidae, Milyeringidae, Eleotridae, Butidae, Thalasseleotrididae, Oxudercidae, and Gobiidae, belong to the suborder Gobioidei, whereas the Trichonotidae, the only family in suborder Trichonotoidei, is the sister group of all species in the suborder Gobioidei (Nelson et al. 2016; Thacker et al. 2015) (Figure 1a). Within Gobioidei, the two families (Oxudercidae and Gobiidae) are the most derived, comprising over 1957 species in 275 genera (Nelson et al. 2016).

FIGURE 1.

Relationships and diversity in Gobiiformes. (a) Phylogeny of Gobiiformes with Kurtiformes as an outgroup. Numbers in parentheses are putative species numbers. The Milyeringidae and Thalasseleotrididae were formerly classified as part of the family Eleotridae, but researchers had noted that these genera were atypical members of the Eleotridae. (b) Photo of Rhyacichthys aspro taken by Dr. Te‐Yu Liao.

Gobioidei are a highly diverse suborder with adaptations to a wide range of environments. Therefore, understanding the genetic basis of these adaptations will shed light on the evolutionary processes that underpin species diversification and the mechanisms that enable such a vast array of species to flourish in both marine and freshwater ecosystems. Access to genome sequences of an organism offers unparalleled insights into its genetic blueprint, supporting the elucidation of its evolutionary history, biological functions, and molecular mechanisms. Moreover, genomic data provide a foundation for comparative studies, helping identify unique adaptations, ecological interactions, and evolutionary innovations. Although many gobioid genome assemblies have been generated recently (Adrian‐Kalchhauser et al. 2020; Cai et al. 2021; Hu et al. 2021; Liu et al. 2021; Yang et al. 2022; You et al. 2014), a reference genome for the earliest divergent family of Gobioidei, the Rhyacichthyidae, is yet to be assembled. Analysis and annotation of a representative Rhyacichthyidae genome will enable comparison to other recently diverged Gobioidei species and should offer invaluable perspectives on the evolution and diversification of this suborder.

Within Gobioidei, the Rhyacichthyidae exhibit several comparatively ancestral characteristics, such as three epurals, six branchiostegal rays, and unfused pelvic fins, that differentiate it from derived relatives (Hoese and Gill 1993; Miller 1973). Perhaps the most distinct feature of Rhyacichthyidae is its sensory system. Lateral‐line systems of the vast majority of gobioid fishes are reduced to cephalic canals, accompanied by a remarkable proliferation of petite superficial neuromasts (located on ‘sensory papillae’). This unique sensory system may indicate that derived gobies have specialised neuron and sensory development. The Rhyacichthyidae, however, do not possess this specialised feature. Instead, they have a well‐developed lateral line that runs along the head and body (Akihito et al. 2000; Hoese and Gill 1993; Miller 1973).

The Rhyacichthyidae comprise only two genera and three species including Rhyacichthys aspro (Valenciennes 1837) (Figure 1b), commonly known as the loach goby, that can be found throughout the Indo‐Australasian archipelago, China, Taiwan, and Japan (Froese and Pauly 2023). Adult loach gobies inhabit freshwater streams of typically rocky or gravelly areas with fast‐flowing water. Morphologically, the species is characterised by an elongated, cylindrical body with a notably flattened head and snout. Loach gobies inhabit benthic environments, primarily consuming diatoms and algae. A previous study based on otoliths suggests that the species is amphidromous (Tabouret et al. 2014) migrating between freshwater and marine environments to complete its life cycle.

To elucidate the genomic basis of species diversification and evolution in Gobioidei, we sequenced, assembled, and annotated a loach goby, R aspro , genome from northern Taiwan. We then performed a comparative genomic analysis using nine published chromosomal scale genomes representing four Gobioidei families (Table 1). Our results show that the expansion of transposable elements (TEs), especially DNA transposons, is associated with an increased substitution rate and species diversification of Gobioidei, suggesting that DNA transposon activity is a key factor in the extensive diversity and adaptability of Gobioidei. Additionally, we identified unique gene sets related to synaptic development, neurocranium morphogenesis, and osmoregulation. These findings illuminate the organism's adaptations to sensory demands, turbulent currents, and the amphidromous nature of its habitats.

TABLE 1.

Genome statistics of the selected species.

Order	Family	Species	Common name	Genome size (Mb)	Classified repeats (%)	Number of scaffolds	N50 (Mb)	L50	BUSCO v5 (%)	Accession
Gobiiformes	Rhyacichthyidae	Rhyacichthys aspro	Loach goby	594	28.2	1489	24	12	98.43	GCA_036850595.1 a
Gobiiformes	Odontobutidae	Perccottus glenii	Chinese sleeper	761	46.5	184	34	10	96.79	GCA_024416835.1
Gobiiformes	Butidae	Bostrychus sinensis	Four‐eyed sleeper	865	46.2	3915	38	11	97.77	GCA_018703735.1
Gobiiformes	Oxudercidae	Boleophthalmus pectinirostris	Great blue‐spotted mudskipper	956	53.9	16620	2	89	96.51	GCF_000788275.1
Gobiiformes	Oxudercidae	Periophthalmus magnuspinnatus	Korean giant‐fin mudskipper	739	51.3	124	33	11	97.14	GCF_009829125.1
Gobiiformes	Oxudercidae	Periophthalmus modestus	Shuttles hoppfish	854	52.6	1419	33	12	95.96	GCA_019594935.1
Gobiiformes	Oxudercidae	Mugilogobius chulae	Yellowstripe goby	1003	53.5	732	42	11	92.12	GCA_016735935.1
Gobiiformes	Oxudercidae	Rhinogobius similis	Barecheek goby	890	54.7	22	42	10	95.55	GCA_019453435.1
Gobiiformes	Gobiidae	Neogobius melanostomus	Round goby	1003	58.4	1364	3	88	94.56	GCA_007210695.1
Gobiiformes	Gobiidae	Proterorhinus semilunaris	Tubenose goby	865	55.1	217	39	10	96.4	GCA_021464625.1
Kurtiformes	Apogonidae	Sphaeramia orbicularis	Orbiculate cardinalfish	1343	60.8	340	57	11	97.69	GCF_902148855.1
Kurtiformes	Apogonidae	Siphamia tubifer	Tuber cardinalfish	1201	59.2	119	51	11	98.02	GCA_020466265.1

^a This study.

2. Materials and Methods

2.1. Sample Collection

An adult female loach goby, Rhyacichthys aspro , was collected from Masu creek (Masu1) (25°10′16.4″ N 121°40′47.7″ E), New Taipei, Taiwan. The head was excised from the fish body and preserved in RNAlater Stabilisation Solution (Cat: AM7021, Invitrogen, Thermo Fisher Scientific Inc.) at −80°C in a freezer. The other tissues were preserved in 100% ethanol at −80°C in a freezer.

To investigate nucleotide diversity and population size of the loach goby in Taiwan, four additional individuals were collected and sequenced. Two individuals (Gangkou1 and Gangkou2) were collected from Gangkou creek, Pingtung, Taiwan, one (JhonggongM) from Jhonggong creek, Miaoli, and one (JhonggongP) from Jhonggong creek, Pingtung (Table S1, Appendix S1). All experiments in this study were performed in accordance with guidelines of the animal ethics committee and were approved by the Institutional Animal Care and Use Committee (IACUC 20‐12‐1593), Academia Sinica.

2.2. DNA and RNA Sequencing

30–50 mg of muscle tissue was lysed with proteinase K and Buffer G2 (with RNAse A) of Genomic DNA Buffer Set (Cat: 19060, Qiagen) and the genomic DNA was extracted via Genomic‐tip 20/G (Cat: 10223, Qiagen), following the manufacturer's instructions.

Genomic DNA (5–8 μg) for Nanopore sequencing was end‐repaired and ligated via a KAPA Hyper Prep Kit (Cat#KR0961, Kapa Biosystems, Wilmington, MA, USA), following the manufacturer's instructions. The genomic DNA library was premixed with the LB and SQB buffer in the Nanopore Ligation Sequencing Kit (SQK‐LSK109, Oxford Nanopore Technologies, UK) and sequenced on a flow cell (R9.4.1; FLO‐MIN106) using MinION software version 19.12.5 for 72 h. Basecalling was performed using the default settings of ONT Guppy software version 3.3.0 ± c1fce59 in high accuracy mode. The resulting median quality score was 18, with a read N50 of 6483 bp.

Genomic DNA was sheared using a Covaris 2 machine, and then quantified using a Qubit 2.0 Fluorometer before Illumina sequencing. A short‐read sequencing library was also constructed from 200 ng of the sheared gDNA using the TruSeq Nano DNA High Throughput Library Prep Kit (Cat. No. 20015965). It was then sequenced using an Illumina X‐ten through the Genomics BioScience & Technology (New Taipei City, Taiwan).

Five to 30 mg of eye, olfactory lobe, gill, liver, muscle, and brain tissues were excised and homogenised with 600 μL Trizol solution for RNA sequencing. RNA was extracted from the homogenised Trizol solution using RNeasy Plus Mini Kit (Cat: 74134, Qiagen), and quantified using the Agilent 2100 Bioanalyser. A short‐read sequencing library was constructed from approximately 1.5 μg of the quantified RNA using the TruSeq Stranded mRNA Library Prep Kit (Cat. No. 20020595). It was then sequenced on the Illumina NovaSeq 6000 through the services of Genomics BioScience & Technology.

2.3. Draft Genome Assembly

Nanopore long reads were assembled using Canu v1.8 (Koren et al. 2017) with ‘‐nanopore‐raw’ parameter. The resulting assembly was error‐corrected using: (1) Racon v1.4.3 (Vaser et al. 2017) via Canu‐corrected Nanopore long reads four times; (2) Medaka v0.11.2 (https://github.com/nanoporetech/medaka) via Canu‐corrected Nanopore long reads once; and (3) pilon v1.23 (Walker et al. 2014) via trimmed Illumina short reads (using Trimmomatic v0.39 with the following parameters: ILLUMINACLIP:TruSeq3‐PE.fa:2:30:10:2:True LEADING:20 TRAILING:20 SLIDINGWINDOW:4:15 MINLEN:36) (Bolger et al. 2014).

2.4. Chicago and HiRise Scaffolding

Both Chicago libraries (Putnam et al. 2016) and Dovetail HiC libraries were prepared as described previously (Lieberman‐Aiden et al. 2009). The de novo draft assembly, Chicago library reads, and Dovetail HiC library reads were used as input data for HiRise, a software pipeline designed specifically for using proximity ligation data to scaffold genome assemblies (Putnam et al. 2016). An iterative analysis was conducted. First, Chicago library sequences were aligned to the draft input assembly using a modified SNAP read mapper (http://snap.cs.berkeley.edu). The separations of Chicago read pairs mapped within draft scaffolds were analysed by HiRise to produce a likelihood model for genomic distance between read pairs, and the model was used to identify and break putative misjoins, to score prospective joins, and to make joins above a threshold. After aligning and scaffolding Chicago data, Dovetail HiC library sequences were aligned and scaffolded following the same method.

2.5. Genome Quality Analysis and Genome Masking

The genome quality was calculated using BUSCO v5 (Manni et al. 2021) score in actinopterygii_odb10. BlobTools v1.1 (Laetsch and Blaxter 2017) was used to check for assembly contamination. The genome size was estimated using Jellyfish (Marcais and Kingsford 2012).

To characterise TEs, a consensus library was built by merging the output from a de novo library from RepeatModeler2.0.2 (Flynn et al. 2020), an LTR library from LTR_finder (Xu and Wang 2007) and LTR retriever (Ou and Jiang 2017), the TE_library from RepBase20181026 (Bao et al. 2015) and Dfam3.5 (Storer et al. 2021), and the P. magnuspinnatus TE library from FishTEDB (Shao et al. 2018). TEs were masked using RepeatMasker 4.1.2 (Nishimura 2000). Kimura divergence was retrieved from RepeatMasker, and the repeat landscape was plotted using the python script ‘TE_divergence_landscape.py’, available on Github (https://github.com/PlantDr430/CSU_scripts/blob/master/TE_divergence_landscape.py).

2.6. Genome Annotation of the Loach Goby

BRAKER 2.1.6 (Hoff et al. 2019) was used for genome annotation. It predicts genes based on RNA or homologous protein sequences. For RNA‐based prediction, transcriptomes from gill, brain, eye, liver, nose, and muscle were mapped to the assembly using hisat2 v2.2.1 (Kim et al. 2019) to create a bam file. BRAKER was run with the genome, RNA bam file, and options ‘‐‐softmasking’ and ‘‐‐UTR = on’.

For homology protein based prediction, gene sets of 12 Percomorphaceae species, including Periophthalmus magnuspinnatus (GCF_009829125.1), Boleophthalmus pectinirostris (GCF_000788275.1), Sphaeramia orbicularis (GCF_902148855.1), Betta splendens (GCF_900634795.3), Scophthalmus maximus (GCF_013347765.1), Oryzias latipes (GCF_002234675.1), Nothobranchius furzeri (GCF_001465895.1), Gasterosteus aculeatus (GCF_016920845.1), Larimichthys crocea (GCF_000972845.2), Syngnathus acus (GCF_901709675.1), Thalassophryne amazonica (GCF_902500255.1) and Takifugu rubripes (GCF_901000725.2) were downloaded from NCBI. An orthology database was created based on genome and protein gene sets using the ProHint BRAKER extension. BRAKER was run with the genome sequence, protein orthology database, and the option ‘‐‐softmasking’. The TSEBRA BRAKER extension was used to combine RNA and protein‐based annotations. The sequences were then blasted against the NCBI database with an E‐value cut off of 1.0E⁻⁵ and any sequences without BLAST hits were filtered out.

Functional annotation, including NCBI BLAST and InterProScan analyses (Jones et al. 2014), was performed using Blast2GO (Conesa et al. 2005). Gene Ontology (GO) annotations were generated with EggNOG‐mapper2.1.0 (Cantalapiedra et al. 2021), employing the EggNOG5.0.2 database (Huerta‐Cepas et al. 2019). BLAST searches were carried out against the Actinopterygii taxonomic group in the NCBI nr database. InterProScan queried multiple databases, including CDD (Marchler‐Bauer et al. 2010), HAMAP (Lima et al. 2009), PANTHER (Mi et al. 2005), Pfam (Bateman et al. 2004), PIR (Barker et al. 2000), PRINTS (Attwood et al. 2012), ProSite (Hulo et al. 2006), TIGRFAM (Haft et al. 2012), Gene 3D (Lees et al. 2012), SFLD (Akiva et al. 2014), and SuperFamily (Lima et al. 2009). Subsequently, KAAS (Moriya et al. 2007) was used to search KEGG Orthology identifiers (KO numbers) in selected species using the bidirectional best hit (BBH) method and mapped them to the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (Kanehisa and Goto 2000). Comparative gene sets in the KEGG database were selected from Homo sapiens , Mus musculus , and all 38 fish species. Finally, PfamScan (Mistry et al. 2007) was utilised to search for protein domains within the predicted genes using the Pfam database (Bateman et al. 2004).

2.7. Comparative Genome Analysis

We downloaded the genomes of related species listed in Table 1. Because different databases have been utilised to identify TEs of different species in the literature, we reanalyzed the TEs of the genomes listed in Table 1 to ensure they are compared using the latest database. Computational Analysis of gene Family Evolution (CAFÉ) 5 (De Bie et al. 2006) was used to calculate the expansion and reduction of the TEs based on copy number within each species and the branch lengths in the phylogeny.

We also annotated the genomes that lacked existing annotations, using the protein‐based method as described in section 2.6. The completeness of these annotations was assessed using the peptide mode of BUSCO v5 (Manni et al. 2021), employing the actinopterygii_odb10 dataset. The functional annotation of these genomes was conducted following the procedures described in section 2.6. The results for GO, Pfam, and KEGG were each transformed into separate copy number distribution matrices. These matrices were then standardised to z‐scores to assess significance. Loach goby with z‐scores > 1.96 were considered significantly enriched, while those with z‐scores < −1.96 were considered significantly reduced.

Orthofinder2.52 (Emms and Kelly 2019) was used to identify orthogroups. Single‐copy orthogroups were selected and aligned using MACSEv2.05 (Ranwez et al. 2011). To minimise potential alignment errors, we excluded genes exhibiting frameshift mutations or premature stop codons. In addition, regions that could not be reliably aligned, specifically those with more than three nucleotide substitutions within two consecutive codons, were removed from further analysis. IQ‐TREE v1.6.12 (Nguyen et al. 2015) was then used to construct the phylogenetic tree. Synonymous (dS) and nonsynonymous (dN) substitution rates were calculated using codeml of PAML4.9 (runmode = 2, seqtype = 1, CodonFreq = 2, fix_omega = 0) (Yang 2007). The pairwise genetic distance of single‐copy nuclear genes and mitochondrial 12S and 16S rDNAs was calculated using MEGA11.0.13 (Tamura et al. 2021).

Syntenic relationships between the loach goby and other selected species ( Periophthalmus magnuspinnatus , P. modestus (GCA_019594935.1), and S. orbicularis ) with chromosome‐level genome assemblies were inferred using DAGchainer (Haas et al. 2004) bundled by Synima (Farrer 2017). The chromosome ordering is based on the assembly of S. orbicularis . GENESPACE was used to visualise chromosome synteny (Lovell et al. 2022).

We categorised TEs inserted in regulatory regions into different groups based on their Kimura distances (0%–10%, 10%–20%, 20%–30% and 30%–40%). To minimise potential confounding effects, we excluded regions containing multiple TE insertions. Additionally, we identified regulatory regions without any TE insertions (no_insert) to serve as a baseline group for comparison.

To investigate the potential regulatory impact of TEs in gene regulatory regions through DNA methylation, we examined the normalised CpG observed/expected (CpG o/e) ratio in TEs located within the 5 kb upstream regions of protein‐coding genes. The normalised CpG fraction was calculated as the ratio of observed CpG to expected CpG, where the expected CpG value was determined using the formula (GC content/2)² (Saxonov et al. 2006).

2.8. Population Genetic Tests and Population Size Estimation

To analyse the nucleotide diversity and estimate the population size of our loach goby population, Illumina short reads of Masu1, Gangkou1, Gangkou2, JhonggongM, and JhonggongP were mapped back to the reference genome using BWA0.7.17 (Li 2013). VCF files were then generated using samtools1.12 and bcftools1.12‐26. The nucleotide diversity pi and Tajima's D statistics were calculated using vcftools with a vcf file merged from five samples using bcftools.

To identify selective sweeps, Tajima's D values were calculated using vcf.‐kit for 5‐kb sliding windows in 1‐kp steps. Genes with a 1 kb flanking region containing at least one window with negative D was considered as positively selected. GO enrichment analysis of positively selected genes was conducted using Blast2GO (Conesa et al. 2005). Only the longest isoform sequences of loach goby were used, and significance was determined using Fisher's exact test with a FDR < 0.05.

For MSMC (Schiffels and Wang 2020) analysis, sites with base quality lower than 20 and mapping quality lower than 20 or mappability rate lower than one were ignored. Read depths lower than one‐third or higher than twice the average were likewise discarded. Diploid fastq was created using vcfutils.pl from filtered vcf files containing data from each sample. Population sizes were then estimated using the resulting fastq. ‐p (The patterns of parameters) was set at ‘1*2 + 50*1 + 1*2 + 1*3’. The generation time was estimated to be 3 years (Li et al. 2024). The mutation rate of R. aspro was inferred to be 4.08 × 10⁻⁹ per nucleotide per generation, based on the known mutation rate of Boleophthalmus pectinirostris (You et al. 2014) and the ratio of R. aspro 's branch length to that of B. pectinirostris .

3. Results

3.1. Genome Assembly and Annotation

We initially assembled a draft genome from 16.7 Gbp of Nanopore and 21.4 Gbp of Illumina reads (Table S1, Appendix S1). We produced 4140 contigs and 607 Mb in total length, with a contig N50 value of 3.1 Mbp (Table S2, Appendix S1). Blobtools analysis indicated that 99.6% (604.4 Mb) of the assembly aligns with the Chordata phylum, whereas 0.4% (2.4 Mb) did not match any known categories (Figure S2, Appendix S2). The estimated genome size based on k‐mer analysis is 604 Mb (Figure S2, Appendix S2), which is consistent with the assembled size attributed to Chordata in the Blobtools analysis. To enhance assembly contiguity, we generated Chicago and Hi‐C libraries using Dovetail Genomics (Scott Valley, California), facilitating the assembly of a 594 Mb genome into 1489 scaffolds. The resulting assembly featured a scaffold N50 value of 24 Mb and organised 90% of the genome into 23 distinct chromosome‐level scaffolds (Figure S2, Appendix S2).

We obtained nine additional Gobioidei genomes from public sources (Table 1) to study Gobioidei evolution. To ensure the quality of further analyses, the selection criterion was that their genomic BUSCO scores must be higher than 90%. The BUSCO score for our assembled loach goby genome was 98.43%, the highest among the species compared (94.56%–98.43%). Gobioidei species genome sizes, ranging from 594 to1343 Mb, appear to increase from early‐diverging to more recently diverged groups (Figure 1a). Notably, the loach goby was placed in the most early‐diverging position and has the smallest genome among the studied species.

BRAKER2.1.6 (Hoff et al. 2019) was used for genome annotation. We annotated 25,259 loach goby genes with 96.24% BUSCO completeness (Table 2 and Table S3, Appendix S1), the majority of which are located on the 23 chromosome‐level scaffolds (Table S2b, Appendix S1). Among these genes, 89.84% match InterProScan, 66.62% have GO terms, 66.06% have orthologs in KEGG, and 48.85% are annotated across all three databases (Figure S3, Appendix S2). On average, each mRNA comprises 10.5 exons that are approximately 160 bp long on average. These data are consistent with other annotated genomes (Table S3, Appendix S1). However, the mean loach goby intron is only 892 bp long, considerably shorter than in other species (1020–2688 bp, p < 10⁻³; Table 2).

TABLE 2.

Statistics of the predicted gene.

Order	Family	Species	Number of genes	Mean protein length (bp)	Mean exon length (bp)	Mean intron length (bp)	BUSCO v5 (%)	Citation
Gobiiformes	Rhyacichthyidae	Rhyacichthys aspro	25259	556	160	892	96.24	This study
Gobiiformes	Odontobutidae	Perccottus glenii	36156	442	168	864	88.46	This study
Gobiiformes	Butidae	Bostrychus sinensis	39495	429	183	1,121	91.04	This study
Gobiiformes	Oxudercidae	Boleophthalmus pectinirostris	22600	587	196	1,963	98.21	You et al. 2014
Gobiiformes	Oxudercidae	Periophthalmus magnuspinnatus	24442	647	202	1,634	98.16	https://www.ncbi.nlm.nih.gov/genome/35290
Gobiiformes	Oxudercidae	Periophthalmus modestus	30505	535	160	1,020	86.40	Yang et al. 2022
Gobiiformes	Oxudercidae	Mugilogobius chulae	53446	315	166	1,229	67.12	This study
Gobiiformes	Oxudercidae	Rhinogobius similis	42619	415	193	1,232	82.69	This study
Gobiiformes	Gobiidae	Neogobius melanostomus	38773	378	171	1,631	70.41	Adrian‐Kalchhauser et al. 2020
Gobiiformes	Gobiidae	Proterorhinus semilunaris	32912	478	185	1,128	81.79	This study
Kurtiformes	Apogonidae	Sphaeramia orbicularis	28580	706	263	2,688	99.09	https://www.ncbi.nlm.nih.gov/genome/82499
Kurtiformes	Apogonidae	Siphamia tubifer	40691	438	181	1,482	89.89	This study

To examine chromosomal evolution in Gobioid fishes, we used DAGchainer (Haas et al. 2004) to deduce synteny relationships between the loach goby genome and other chromosome‐level assembled genomes, including Periophthalmus magnuspinnatus , P. modestus , and Sphaeramia orbicularis . We see one‐toto‐one chromosome correspondence among these species, with the exception of chromosomes 12 and 23 of P. magnuspinnatus that match chromosome 12 in the other species. Thus, despite considerable genome size variability, Gobioid fish chromosome numbers remain remarkably conserved (Figure 2, Table S4, Appendix S1).

FIGURE 2.

Syntenic chromosome conservation in gobioid fishes. The synteny relationships among Rhyacichthys aspro (R. asp), Periophthalmus magnuspinnatus (P. mag), P. modestus (P. mod) and Sphaeramia orbicularis (S. orb) revealed the conservation of chromosome numbers among gobioid fishes. Chromosomal orders referenced here are based upon the configuration of S. orbicularis . It is noteworthy that what is identified as chromosome 12 and 23 in P. magnuspinnatus correspond to what is designated as chromosome 12 in other species. Additionally, S. orbicularis possesses only 23 chromosomes, lacking a distinct chromosome 23; instead, its chromosome 24 is homologous to chromosome 23 of R. aspro and P. modestus . The synteny relationships were visualised using GENESPACE.

We identified 2100 single‐copy orthologs in the 10 sampled gobioid and two Kurtiformes genomes using Orthofinder (Emms and Kelly 2019). These orthologous genes were concatenated and aligned to construct a maximum‐likelihood phylogenetic tree (Figure 3). The phylogeny is consistent with previous studies, showing that Rhyacichthyidae and Odontobutidae together form the early‐diverging clade in Gobioidei (Agorreta et al. 2013; Thacker et al. 2023; Tornabene et al. 2013) followed by Butidae. Oxudercidae and Gobiidae are sister clades and are the most derived groups. Molecular dating was performed using MCMCtree (Yang 2007) and incorporated data obtained from TIMETREE (Kumar et al. 2022). The divergence between Gobiiformes and Kurtiformes was estimated at around 88 million years ago (MYA; 95% interval 86–92 MYA), at the lower end of previous estimates between 86 and 102 MYA (Alfaro et al. 2018; Hughes et al. 2018; Rabosky et al. 2018).

FIGURE 3.

Phylogeny of Gobioidei. The phylogenetic tree was constructed using maximum likelihood methods in IQ‐TREE, based on 2100 single‐copy orthologs, with the Kurtiformes serving as the outgroup. All nodes are supported by > 90% bootstrap values. Adjacent to the nodes, numbers represent the divergence times in millions of years, with the 95% confidence within parentheses. Below the branches, numbers indicate genetic distances, while gains (in blue) and losses (in red) of transposable elements are shown above the branches.

3.2. Repetitive Elements and Gobioid Evolution

Mirroring the trend in genome size, TEs increase from early‐ to recently‐diverged Gobioidei groups. Notably, the loach goby genome has the fewest TEs among the species analysed (Table 1), a pattern consistent across all TE types (Table S5a, Appendix S1). Genome sizes strongly correlate with TE abundance (ρ = 0.85, p < 0.01) after correcting for phylogenetic relationships using phylogenetically independent contrasts (PIC) (Felsenstein 1985) (Figure S4a, Appendix S2). Excluding TEs strikingly reduces genome size variability, indicating that TE changes predominantly drive genome size variation (Figure S4b, Appendix S2). We analysed TE dynamics across the Gobioidei phylogeny using CAFÉ (Figure 3) and found pronounced TE expansion in lineages leading to recently diverged groups as well as in the outgroup Kuritiformes. This supports the hypothesis that ancestral Gobioidei had fewer TEs, suggesting that the smaller genomes in early‐diverging groups mainly result from TE expansion in recently diverged taxa, not from TE loss during evolution.

According to the repeat landscapes (Figure 4), the loach goby genome shows the least TE expansion compared to others. Percccotthus glenii (Odontobutidae) exhibits limited historical TE expansion with increased recent activity (divergence level < 5%) (see also Figure 3). Butidae and Oxudercidae have experienced a burst of TE expansion leading to a high prevalence of among‐locus divergence between 10% and 20%. In contrast, Gobiidae species show more recent TE activity (divergence level < 10%). In all cases, DNA transposons dominate TE expansion dynamics.

FIGURE 4.

Comparison of the repeat landscapes among the 12 studied genomes. Note that the majority of expanded transposable elements are DNA transposons.

Figure 3 shows that branches leading to clades harbouring few TEs, including Rhyacichthyidae, Odontobutidae, and Butidae, are shorter than those leading to TE‐rich groups (Oxudercidae and Gobiidae). Indeed, the loach goby branch is the shortest. To understand this, we analysed genetic distances between Gobioidei and the Kurtiformes outgroup (Table S6, Appendix S1), using these distances as proxies for substitution rates. These discrepancies highlight differential substitution rates post‐divergence. We found a strong positive correlation of genetic distance and TE prevalence (ρ = 0.87, p < 0.01, corrected for phylogeny using PIC; Figure 5), including both synonymous (ρ = 0.75, p < 0.02; Figure S5a, Appendix S2) and nonsynonymous sites (ρ = 0.9, p < 0.01; Figure S5b, Appendix S2). Separate analysis of TE types shows similar correlations for DNA transposons (ρ = 0.85, p < 0.01) and unclassified TEs (ρ = 0.87, p < 0.01), but not for retrotransposons (Figure S6, Appendix S2). This correlation extends to different DNA transposon superfamilies, particularly hAT (ρ = 0.68, p = 0.05) and hATAc (ρ = 0.73, p = 0.03; Table S5b, Appendix S1, Figure S7, Appendix S2). DNA repair after transposon excision and insertion may increase mutation rates in adjacent sequences (Wicker et al. 2016). Beyond TE abundance, we assessed the cumulative alterations from TE activity (insertions and excisions), finding strong correlations with genetic distances for all DNA transposons (ρ = 0.92, p < 0.01) and hAT transposons in particular (ρ = 0.87, p < 0.01; Figure S7, Appendix S2). However, mitochondrial gene distances showed no correlation with TE content (Table S7, Appendix S1, Figure S8, Appendix S2).

FIGURE 5.

Correlation between different transposable element types and substitution rate. The genetic distances were distance to outgroup Kurtiformes. Since all Gobioidei were derived from the outgroups at the same time, the genetic distances served as a proxy for substitution rate. The correlation was conducted using Kendall's ρ coefficient after correcting for phylogeny using phylogenetically independent contrasts (see text for details).

We analysed TE prevalence in genic (coding + intron) and regulatory regions (within 5Kb upstream of translation start site), as TE within these regions can directly impact gene function and phenotypic diversity (Chuong et al. 2017; Fotsing et al. 2019; Shi et al. 2023). DNA transposons, particularly hAT and hAT‐Ac types, were found to be more abundant than retrotransposons (SINEs, LINEs, and LTRs) in these areas (p < 0.01; Figure 6). Although percentage differences seem marginal, the actual count of DNA transposons significantly exceeds that of retrotransposons, indicating substantial variations in actual counts (Table S5a, Appendix S1). This pattern is less evident in Kurtiformes. To examine TE activity historically, we categorised them based on insertion times, as measured by Kimura distances, into three ranges: 0%–10%, 10%–20% and 20%–30%. Variations in these categories indicate changes in TE activity over time. The analysis reveals that DNA transposons are consistently more abundant in genic and regulatory regions (Figure S9, Appendix S2), suggesting that they have been among the most active TEs in the Gobioidei.

FIGURE 6.

The proportion of various types of transposable elements in functional regions. The functional regions include genic (coding + intron) and the regions within 5 kilobases (Kb) directly upstream of the translation start site. Differences among various types of TEs were assessed using the Wilcoxon rank‐sum test.

Due to the potentially harmful effects of ectopic expression of TEs, hosts have evolved CpG site methylation as a mechanism to suppress their transcriptional activities (Zhou et al. 2020). We examined the potential for DNA methylation across TEs of different age using CpG observed/expected (CpG o/e) ratios as indexes. We found that regions with recently inserted TEs (0%–10% of divergence) exhibit higher CpG index compared to those with more ancient insertions (30%–40% of divergence) or no TE insertions (Figure S10, Appendix S2). This pattern underscores the impact of TE insertion on the CpG index in these regions, suggesting that methylation at CpG sites is crucial in suppressing TE activity.

3.3. Genes Related to the Loach Goby Natural History

Our analyses of the loach goby genome also shed light on the special features of this species (Table 3 and Tables S8–S13, Appendix S1). Tables S8–S13 listed the expanded/reduced GO term (Tables S8–S10, Appendix S1), protein family (Table S11, Appendix S1), protein domain (Table S12, Appendix S1) and KEGG number (Table S13, Appendix S1). In these analyses, Mugilogobius chulae and Neogobius melanostomus were excluded due to their BUSCO values being below 80%. Additionally, P. modestus was omitted in favour of its congeneric, P. magnuspinnatus , which has a higher BUSCO value, to avoid overrepresentation of the genus in the analysis.

TABLE 3.

Significantly enriched/depleted GO terms biological process in Rhyacichthys aspro .

GO_ID	GO_term	R. asp	P. gle	B. sin	B. pec	P. mag	R. sim	P. sem	S. orb	O. lat	G. acu	T. rub	z‐score
GO:0048920	Posterior lateral line neuromast primordium migration	3	4	4	4	4	4	4	4	4	4	4	−3.02
GO:0035676	Anterior lateral line neuromast hair cell development	0	1	1	1	1	1	1	1	1	1	1	−3.02
GO:0048901	Anterior lateral line neuromast development	0	1	1	1	1	1	1	1	1	1	1	−3.02
GO:0048903	Anterior lateral line neuromast hair cell differentiation	0	1	1	1	1	1	1	1	1	1	1	−3.02
GO:0016080	Synaptic vesicle targeting	75	46	56	27	27	57	46	29	27	29	31	2.09
GO:0051601	Exocyst localization	61	26	23	23	20	24	18	17	17	17	18	2.92
GO:0001927	Exocyst assembly	45	10	10	5	7	10	9	5	5	4	6	2.96
GO:0048702	Embryonic neurocranium morphogenesis	16	15	15	14	14	14	14	14	13	13	15	1.91*
GO:0072206	Metanephric juxtaglomerular apparatus development	3	2	2	2	2	2	2	2	2	2	1	2.24
GO:0072227	Metanephric macula densa development	3	2	2	2	2	2	2	2	2	2	1	2.24
GO:0072240	Metanephric DCT cell differentiation	4	3	3	3	3	3	3	3	3	3	2	2.24
GO:0072221	Metanephric distal convoluted tubule development	7	3	6	3	4	3	4	3	3	3	2	2.20
GO:0072233	Metanephric thick ascending limb development	6	2	5	2	3	2	3	2	2	2	1	2.20
GO:0072218	Metanephric ascending thin limb development	8	5	7	4	5	5	5	4	4	4	3	2.14
GO:0072235	Metanephric distal tubule development	9	6	8	5	6	6	6	5	5	5	4	2.14
GO:0072051	Juxtaglomerular apparatus development	3	2	2	2	2	2	2	2	2	2	1	2.24
GO:0072024	Macula densa development	3	2	2	2	2	2	2	2	2	2	1	2.24
GO:0072051	Juxtaglomerular apparatus development	3	2	2	2	2	2	2	2	2	2	1	2.24
GO:0072069	DCT cell differentiation	4	3	3	3	3	3	3	3	3	3	2	2.24
GO:0072023	Thick ascending limb development	6	2	5	2	3	2	3	2	2	2	1	2.20
GO:0072021	Ascending thin limb development	8	5	7	4	5	5	5	4	4	4	3	2.14

Abbreviations: B. pec, Boleophthalmus pectinirostris ; B. sin, Bostrychus sinensis ; G. acu, Gasterosteus aculeatus ; O. lat, Oryzias latipes ; P. gle, Perccottus glenii ; P. mag, Periophthalmus magnuspinnatus ; R. asp, Rhyacichthys aspro ; R. sim, Rhinogobius similis ; S. orb, Sphaeramia orbicularis ; T. rub, Takifugu rubripes .

^*Becoming significant (z‐score = 1.98) when the analysis was restricted to include only Gobiiformes and its sister group Kurtiformes.

Specifically, the loach goby has fewer genes associated with neuromast development and differentiation than other Gobioidei species, including genes involved in posterior lateral line neuromast primordium migration (GO:0048920) (Table S14a, Appendix S1) and anterior lateral line neuromast development and differentiation (GO:0035676; GO:0048901, GO:0048903) (Table S14b, Appendix S1). These differences may contribute to the unique mechanosensory lateral line system observed in the loach goby (see Introduction). Some genes in these GO terms are known to be involved in neuron system development and function. For example, sphingosine‐1‐phosphate receptor 2 (s1pr2) (Table S14a, Appendix S1) has an important role in the development of otic vesicles, neuromasts, and survival of hair cells. Misregulation of s1pr2 leads to a change in hair cell and neuromast abundance and their lateral line positioning in zebrafish (Hu et al. 2013). The insulinoma‐associated protein 1a (insm1a) (Table S14b, Appendix S1) is a zinc‐finger transcription factor playing a variety of roles in cell formation and differentiation of vertebrate central and peripheral nervous systems (Gong et al. 2017). Knockdown of insm1a expression in zebrafish leads to defects in primordium migration, resulting in fewer neuromasts (He et al. 2017).

Interestingly, the loach goby genome also has an overabundance of genes involved in synaptic formation and in sensory information processing, including synaptic vesicle targeting (GO:0016080), exocyst localization (GO:0051601) and assembly (GO:0001927). Consequently, while the loach goby has an ancestral lateral line system, it might have evolved unique sensory perception and lateral line development pathways at the molecular level.

Morphologically, loach goby is characterised by a flattened head and abdomen, which may be critical for adaptation to torrential currents. It is thus notable that the genome of this species has a higher number of genes associated with embryonic neurocranium morphogenesis (GO:0048702) (Table 3 and Table S14c, Appendix S1), suggesting a genetic foundation for these morphological adaptations. For example, we find an additional copy of the protein transport factor Sec23A and transforming growth factor beta‐2 proprotein (Tgfb2). Mutations in Sec23A are associated with facial dysmorphisms and late‐closing fontanels in humans (Boyadjiev et al. 2011). Tgfb2 is a precursor of transforming growth factor β‐2 (TGF‐β‐2) that regulates fetal rat cranial suture morphogenesis (Opperman et al. 2000).

Loci involved in metanephric distal tubule (GO:0072235) and metanephric ascending thin limb development (GO:0072218) may underlie the loach goby life history traits. Loach goby larvae drift out into the open ocean and return to fresh water only after becoming juveniles. Therefore, osmoregulation is expected to be particularly important. Some genes, such as POU Domain Class 3 Transcription Factor 3 (POU3F3) and deleted in malignant brain tumours 1 (DMBT1) (Table S14d, Appendix S1), in these GO categories are involved in kidney development and osmoregulation. For example, an amino acid mutation in mouse POU3F3 causes reduced nephron number and impaired kidney development (Rieger et al. 2016). Deletion of DMBT1 results in distal renal tubular acidosis, a kidney condition that impairs the removal of acids from blood into urine (Gao et al. 2010).

To identify genes that might underlie the loach goby adaptive evolution, we used Tajima's D test to identify selective sweeps. We find 1296 genes with negative Tajima's D values (Table S15, Appendix S1). Several GO categories are overrepresented among them (Table S16, Appendix S1). In particular, synapse organisation (GO:0050808) and calmodulin binding (GO:0005516) may influence sensory perception. The organisation of synapses plays a pivotal role in transmitting and processing sensory information across neural circuits (Balaskas et al. 2019). Calmodulin plays a significant role in sensory systems, including vision, hearing, taste, and touch, by mediating the signalling pathways involved in detecting and processing sensory stimuli (Fain et al. 2010; Gillespie and Muller 2009; Wu et al. 2017; Zhang et al. 2003). Using Phenotype Enrichment Analysis (Table S17, Appendix S1) (Weng and Liao 2017), we also found that genes putatively under selective sweeps are disproportionally involved in cranial cartilage development. This finding, coupled with the observed increase in the number of genes associated with embryonic neurocranium morphogenesis (GO:0048702) (Table 3) in the loach goby genome, implies a genetic basis for the species' distinctive flattened head and abdomen. These morphological traits are likely adaptive and facilitate survival in the fast‐flowing waters of their native habitat.

3.4. Population Demography

To understand the genetic diversity and population demography of loach goby, we also generated four additional loach goby genomes (Table S18, Appendix S1). Their heterozygosity ranges from 0.34% to 0.37% with an average nucleotide diversity π of 4.12 × 10⁻³ (or one heterozygous single nucleotide polymorphism every 233 bp).

Our analysis using Multiple Sequentially Markovian Coalescent 2 (MSMC2) (Schiffels and Wang 2020) suggests that loach goby populations have experienced an expansion around 5 × 10⁵ years ago (Figure 7a). The population size grew to 4 × 10⁵ approximately 10⁵ years ago, and then to 1 × 10⁶ by 10⁴ years ago. However, it is worth noting that the MSMC2 tends to be unstable with large haplotype sizes (hap = 10) and may not provide accurate estimates of recent history (Liu and Fu 2020). Therefore, we also utilised Stairway Plot 2 for a more recent demographic estimation (Liu and Fu 2020). Our estimates suggest that loach goby populations began to grow around 6 × 10⁵ years ago, with the size hovering around 4 × 10⁵ between 400,000 and 10,000 years ago (Figure 7b). The populations then experienced a decline, diminishing to an estimated 100,000 individuals, which corresponds to its current population size.

FIGURE 7.

Population demography of loach goby by MSMC2 (a) and stairway plot 2 (b). The demography was based on the genome sequences of five loach gobies.

This recent demographic decline is strongly associated with sea‐level change. The peak in loach goby numbers coincided with lower sea levels and declined as levels rose at the end of the last glacial period, about 10,000 years ago. These patterns may also be linked to changes in habitat availability, since higher sea levels probably shrink freshwater habitats required by the species in their juvenile and adult stages.

4. Discussion

4.1. The Loach Goby Genome Elucidates Gobioidei Diversification

The loach goby genome is small and has a low substitution rate, likely due to a relatively light TE load. While chromosome numbers are remarkably conserved (Figure 2) across Gobioidei species, TE content varies widely, ranging from 28% to 60% of the genome. There is a marked increase in both genome size and TE, particularly DNA transposon prevalence, as we go from early‐diverging to more recently diverged taxa. DNA transposons account for 12% to over 26% of the genomes and 30 to 50% of TEs. They also predominate among expanding classes of repeats (Figure 4). Furthermore, these expansions occur mostly in the species‐rich and derived families, suggesting a positive correlation between TE proliferation and species diversification.

We also find that DNA transposons, especially the hAT and hAT‐Ac families, are disproportionately resided in genic and regulatory regions (Figure 6), consistent with previous studies (Munoz‐Lopez and Garcia‐Perez 2010; Wicker et al. 2016; Woodard et al. 2012). We further detect a strong positive correlation between transposon activity (insertion and excision) and the level of divergence, suggesting that TE movement causes the elevated mutation rates in adjacent sequences (Wicker et al. 2016). Since these transposons are prevalent in functional regions, their mutations likely lead to new functional variants, thereby enhancing the functional complexity required for adaptation to diverse environmental conditions (Chuong et al. 2017; Fotsing et al. 2019; Shi et al. 2023).

In addition, we also observed gradual depletion of CpG sites in these TEs. Interestingly, it has been shown that DNA methylation may enable TEs to acquire new regulatory roles within the host genome. Over time, these once parasitic sequences can become integral parts of the genomic regulatory architecture, contributing to the formation of new regulatory elements like enhancers, which are crucial for gene expression regulation (Zhou et al. 2020). Therefore, the rapid proliferation of DNA transposons may be one of the key contributing factors to the rapid diversification observed within Gobioidei.

TE activity can also facilitate speciation by causing large‐scale genomic variation that creates interspecific genetic incompatibilities (Serrato‐Capuchina and Matute 2018). High TE prevalence in mammalian genomes is associated with increased speciation rate (Ricci et al. 2018). It is thus possible that the diversification of Gobioidei, one of the most species‐rich vertebrate orders, was likely facilitated by the accelerated modification and expansion of TEs in their genomes (Figure 8), promoting novel gene combinations, duplications, and losses, as well as altered gene functions and regulatory networks. These genome‐wide changes, in turn, may have enhanced the capacity of the Gobioidei to adapt to diverse aquatic ecosystems.

FIGURE 8.

Schematic illustration of Gobioidei evolution. Within this suborder, there is a marked increase in both genome size and transposable elements, particularly DNA transposons, from early‐diverging to more recently diverged species. The expansion of DNA transposons is linked to increased mutation rates which may amplify functional complexity for adaptation to varied environments and drive speciation through extensive genomic variations that establish genetic incompatibilities between species. This expansion, notably within functional regions, likely contributes to the remarkable diversification and ecological adaptation observed in this diverse group of vertebrates.

4.2. Genomic Signatures of Adaptation

Analyses of the loach goby genome also reveal special adaptive features of this species. The genome is relatively poor in genes associated with lateral line and neuromast development, but is rich in loci related to synaptic development (Table 3). This suggests that despite the presence of the lateral line system that is ancestral among Gobioidei, this species may have evolved a unique molecular approach in sensory perception and lateral line development. Genes putatively under positive selection are also disproportionately involved in synapse organisation (GO:0050808) and calmodulin binding (GO:0005516) (Table S16, Appendix S1), also likely related to sensory perception (Balaskas et al. 2019; Fain et al. 2010; Gillespie and Muller 2009; Wu et al. 2017; Zhang et al. 2003).

The loach goby genome has copies of Sec23A and Tgfb2 (Table S14c, Appendix S1) not present in related gobioid species. These loci are involved in embryonic neurocranium morphogenesis (GO:0048702) (Table 3). Mutations in these genes affect cranial development and suture morphogenesis (Boyadjiev et al. 2011; Opperman et al. 2000). Cranial cartilage development is also a category overrepresented among putatively adaptively evolving genes (Table S17, Appendix S1). These loci may be associated with flattened heads and abdomens that characterise loach gobies, phenotypes that may underlie adaptation to strong currents. Interestingly, we also found that the genome of a hill‐stream loach, Beaufortia kweichowensis , from the Balitoridae (Cypriniformes) family, which exhibits a flattened cranium adapted to fast mountain streams (Deng et al. 2021), has a higher number of genes in the GO: GO:0048702 category (Table S19a, Appendix S1). The B. kweichowensis genome contains a paralog of Sec23A and Sec23B, which is also involved in embryonic neurocranium morphogenesis (Lang et al. 2006).

Genes related to renal development and osmoregulation (Table 3 and Table S14d, Appendix S1) are also prevalent in the loach goby genome, indicating a molecular mechanism of adaptation to its amphidromous life history. To demonstrate that changes in osmoregulation are particularly important in fish that move between marine and freshwater environments, we also analysed the genome of the European eel ( Anguilla anguilla ), which is a typical catadromous fish (van Ginneken and Maes 2005). Its genome also shows a higher number of genes involved in osmoregulation (Table S19b, Appendix S1).

5. Conclusion

The loach goby genome suggests that TE dynamics may contribute to adaptation and diversification of the Gobioidei, despite the conservation of chromosome number (Figure 8). Specifically, the expansion of DNA transposons is associated with increased substitution rates, likely contributing to the rich diversity of the suborder. Although we have not explicitly determined whether TE expansion is adaptive, the strong correlation we observed underscores its significant impact on genetic variation, which may provide a foundation for evolutionary adaptability in these organisms. Given that the adaptive role of TE expansion remains an open question, it presents an intriguing avenue for future research. Our results suggest that TE‐driven genetic alterations could be a driving force behind the speciation and ecological success of one of the most species‐rich vertebrate orders, particularly in their adaptation to various aquatic ecosystems.

The loach goby genome also points to molecular mechanisms underlying unique sensory and morphological traits. The presence of an expanded repertoire of genes associated with synaptic development and sensory perception, alongside specific gene duplications related to cranial structure, underscores the loach goby's evolutionary response to its ecological niche. These findings not only improve our understanding of the adaptive strategies employed by this species, but also underscore the broader evolutionary mechanisms that enable species to thrive in diverse and changing habitats. Comparative genomic analyses also reveal parallel evolution across species, reinforcing the importance of genetic adaptation through osmoregulation for aquatic species with complex life histories.

Author Contributions

Designed research: T.‐Y.W., S.‐M.C. and H.‐Y.W.; performed sampling: T.‐Y.W., T.‐Y.L., S.‐P.H. and F.‐Y.W.; performed molecular experiment: T.‐Y.W. and F.‐Y.W.; performed bioinformatic analyses: H.‐J.L., Y.‐W.W., J.‐H.T., T.‐H.Y. and H.‐Y.W.; visualisation: H.‐J.L., T.‐H.Y., C.‐T.T. and H.‐Y.W.; statistical analyses: H.‐J.L., J.‐H.T. and H.‐Y.W.; supervision: T.‐Y.W., C.‐T.T., S.‐M.C. and H.‐Y.W.; writing – original draft: T.‐Y.W., H.‐J.L., Y.‐W.W., S.‐M.C. and H.‐Y.W.; and writing – reviewing and editing: T.‐Y.W., H.‐J.L., Y.‐W.W., S.‐P.H., F.‐Y.W., C.‐T.T., S.‐M.C. and H.‐Y.W.

Ethics Statement

All animal experiments in this study were performed in accordance with the guidelines of the animal ethics committee and were approved by the Institutional Animal Care and Use Committee (IACUC 20‐12‐1593), Academia Sinica.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Appendix S1.

Appendix S2.

Data Availability Statement

The loach goby ( R. aspro ) genome project was deposited at NCBI under BioProject No. PRJNA1010998. The Illumina DNA sequencing data were deposited under NCBI accession number SRR28458574; the Nanopore DNA sequencing data were deposited under NCBI accession number SRR28439223; the Hi‐C sequencing data were deposited under NCBI accession number SRR28520832, and the Chicago sequencing data were available under NCBI accession number SRR28520777. The Illumina RNA sequencing reads for different tissues were available under accession numbers SRR28461528 (eye), SRR28466848 (nose), SRR28477822 (brain), SRR28462307 (gill), SRR28466864 (muscle) and SRR28466849 (liver). The genome assembly with gene and transcript annotations has been deposited into GenBank under the accession number GCA_036850595.1 (ASM3685059v1). The Illumina DNA sequencing data of the four additional individuals were available under accession numbers SRR28520831 (Gangkou1), SRR28520833 (Gangkou2), SRR28520835 (JhonggongM) and SRR28520837 (JhonggongP). The accession numbers for other goby species whose genomes were used in this analysis are presented in Table 1. The scripts used in this study, including annotations for genes and TE, can be downloaded from github at the following link (https://github.com/ala98412/LaochGoby_script).