Silver chimaera genome highlights lineage-specific sex chromosome and opsin gene evolution in holocephalans

Abstract

Cartilaginous fishes are divided into holocephalans and elasmobranchs, and they offer valuable systems for analysing the genetic basis of adaptation to diverse habitats and the evolution of chromosomal organization. Genomic studies on cartilaginous fishes were initiated early with holocephalans because of their compact genomes, but have concentrated primarily on the family Callorhinchidae. Here, we focused on the most species-rich holocephalan family Chimaeridae and characterized the genome of its member, silver chimaera (Chimaera phantasma), in pursuit of genomic traces of adaptation to deep-sea vision. The resulting genome assembly exhibited high continuity and completeness, enabling the first chromosome-level comparison among holocephalans. They displayed substantial intragenomic variation in chromosome length, correlated with intron size, alongside a high degree of one-to-one chromosomal homology. Our search for silver chimaera photoreceptor genes revealed a shrunken set of opsin genes, including rhodopsin exhibiting a sequence signature typical of deep-sea adaptation. We also performed whole-genome resequencing of multiple silver chimaera individuals of both sexes, which identified a putative X-chromosome fragment. This is the first evidence of a holocephalan sex chromosome and suggests male heterogametic sex determination. Our findings contribute to a deeper understanding of vertebrate genome diversity and lay the groundwork for future genetic studies on this species.

1. Introduction

Cartilaginous fishes consist of holocephalans and elasmobranchs and provide valuable systems for investigating the genetic basis of adaptation to diverse environments, including the deep sea. While elasmobranchs, consisting of sharks and batoids (skates, rays, torpedoes, and sawfishes), have been repeatedly studied for molecular signatures of visual adaptation to dim-light habitats,^1,2 comparable insights into holocephalans remain limited, with no prominent examples of such adaptations reported to date.³ Holocephalans (subclass Holocephali) are further classified into three families: Chimaeridae, Rhinochimaeridae, and Callorhinchidae (Fig. 1a; reviewed in⁵). Among holocephalans, Callorhinchus milii, often called the elephant shark, was the first species used for the study of visual pigment composition (see above) as well as whole genome sequencing.⁶ For over a decade, this species remained the sole holocephalan with a sequenced genome^7,8 until that of the rabbitfish (Hydrolagus affinis) became available in 2020 (Table 1).⁹

Fig. 1.

Chimaera phantasma. a) Schematic phylogeny. Parentheses include the numbers of described species as of May 2025.⁴ b) Adult individual of Chimaera phantasma. Photo credit: Akinori Teramura. c) Sampling sites. Red circles indicate the location of sample collection.

Table 1.

Basic statistics of Chimaera phantasma genome assemblies and comparison with other holocephalan genome assemblies.

Species	Chimaera phantasma	Callorhinchus milii	Hydrolagus affinis
Assembly ID (Accession ID)	sChiPha1.1	IMCB_Cmil_1.0 (GCF_018977255.1)	UP_Haf (GCA_012026655.1)
Sequencing technology	PacBio	PacBio	Illumina
	CLRa	CLRa	short reads
Assembly programme	wtdbg2	Falcon	W2RAP v. 1.0
Proximity ligation method	iconHi-C	Dovetail Chicago, Dovetail Hi-C	Not used
Scaffolding programme	3d-dna	HiRise	Not used
Total number of bases (Mb)	1,010	941	1,114
Single copy ortholog completenessb	96.8	93.1	56.4
Max. scaffold length (Mb)	155.4	139.2	0.397
N50 scaffold length (Mb)	42.7	69.3	0.019
% of bases in scaffolds > 1 Mb	96.0	90.8	0
Number of sequences > 1 Mb	38	31	0
% Gaps	0.04	0.01	0

^aCLR, continuous long read.

^bCompleteness was measured with the programme pipeline compleasm.¹⁰

The silver chimaera (Chimaera phantasma) (Family Chimaeridae; Fig. 1a) is a demersal deep-sea chimaera that lives in the North Pacific. Because most species in this family are deep-sea dwellers, and they can provide new insights into their ecological adaptation, as previously shown for other chondrichthyans.¹¹ Their deep-sea life makes this group of species highly elusive, and thus it is not trivial to secure tissue samples for the extraction of high molecular weight DNA and intact RNA. However, in some regions of Japan, Ch. phantasma dwells in relatively shallow waters at a depth of approximately 200 m and is more accessible. Chimaera phantasma and Ca. milii, with the above-mentioned long-standing whole genome sequence data, are estimated to have diverged approximately 187 million years ago,¹² and this deep divergence raises the possibility of substantial differences in their karyotypes. Cytogenetic analyses identified highly variable karyotypes within Chimaeriformes: the spotted ratfish (Hydrolagus colliei) has a diploid chromosome number of 58, while the rabbit fish (Chimaera monstrosa) has a diploid chromosome number of 86.¹³ Still, the variation between these karyotypes is largely attributed to the number of tiny chromosomes, so-called ‘microchromosomes’, which are reminiscent of those characterized earlier in birds, reptiles, and some non-teleost actinopterygian fish species.¹⁴ Although no uniform definition of ‘microchromosomes’ has been established (discussed in¹¹), they exhibit remarkable characteristics including higher gene density and higher GC-content.^8,15 Properties of individual chromosomes, however, remain to be further examined in terms of their sequence composition. To address this, holocephalans with chromosomes of variable lengths in their karyotypes offer a valuable study system.

Holocephalan species have relatively small genomes among cartilaginous fishes¹⁶ —none of the five entries found in the Animal Genome Size Database¹⁷ exceeds 2 Gb (mean ± SD = 1.65 Gb ± 0.22). This is far exceeded by the genome sizes of sharks (6.57 Gb ± 3.45) and batoids (4.75 Gb ± 1.75),^18,19 albeit the genome size is still undocumented for hundreds of cartilaginous fish species. Was this remarkable difference in genome size between holocephalans and elasmobranchs driven by variations in ploidy levels, or did it arise independently of whole genome duplication events? Recent studies on elasmobranch genomes have not identified any whole genome duplications in the elasmobranch lineage.^20,21 Therefore, elucidating the mechanisms underlying dynamic genome size changes requires a detailed analysis subdividing the genome into different chromosomes and different segments of genic regions. Chromosome-scale genome assemblies offer a robust foundation for such investigations.

Another important but underexplored aspect of holocephalan biology is the sex determination mechanism. This research direction is driven by efforts to identify sex chromosomes, a task made feasible by high-quality chromosome-level genome assemblies. In 2023, the first chondrichthyan sex chromosomes were reported from zebra shark (Stegostoma tigrinum) and whale shark (Rhincodon typus); in both species, the X chromosome was identified.¹¹ This report was followed by Y chromosome identification in the white-spotted bamboo shark (Chiloscyllium plagiosum).²² On the other hand, sex chromosomes of holocephalans have never been documented. It is worth examining whether holocephalans have a different organization of sex chromosomes from elasmobranchs, after 400 million years of their divergence.

In this study, we sequenced the whole genome of Ch. phantasma and used it to characterize the process of genome evolution leading to the unique karyotypic organization and small genome size in the holocephalan lineage. Our vertebrate-wide comparison of opsin gene repertoires suggests variable retention of different opsin subtypes, possibly associated with different habitat depths. We also report the first genetic evidence of male heterogamety in a holocephalan species, with the identification of a 500 Kb-long putative X chromosome sequence.

2. Methods

2.1. Sample collection

Eleven Ch. phantasma individuals were collected from five sites. The details of the sampling localities used for each analysis are as follows: for long-read sequencing, a young female Ch. phantasma, whose total length (TL) was about 30 cm, caught in Suruga Bay, Shizuoka (Fig. 1b) in 23 January 2020 as a commercial trawl bycatch (Voucher specimen KPM-NI 56434); for RNA sequencing, a young female Ch. phantasma, whose TL was approximately 40 cm, was caught in Suruga Bay, Shizuoka (Fig. 1b), on 20 September 2020, as a commercial trawl bycatch; for short-read genome resequencing, nine males and nine females were collected from the East China Sea, off Okinawa to off Chiba between 15 September 2015, and 12 September 2021, as commercial trawl and longline fishing bycatch (Fig. 1c; Table S1).

We complied with the “Guidelines for the use of fishes in research” published by the Ichthyological Society of Japan in 2003; our samples were limited to the minimum numbers required by the study, and primarily consisted of individuals that were no longer caught as bycatch or had been sold for human consumption. Genomic DNA was extracted from a piece of the right pelvic fin or muscle near the first dorsal fin base and preserved in 99% ethanol.

2.2. Genome sequencing and assembly

Chimaera phantasma was used to obtain high-molecular-weight DNA using the phenol/chloroform method with TNES-urea buffer.²³ The concentration of the extracted DNA was measured using a Qubit Fluorometer (Thermo Fisher Scientific, MA, USA), and the size distribution of the DNA fragments was analysed using a TapeStation 2100 (Agilent Technologies, CA, USA) to ensure high integrity. An SMRT sequence library was constructed with an SMRTbell Express Template Prep Kit 2.0 (Pacific Biosciences, CA, USA) and sequenced in a single 8 M SMRT cell on a Sequel II system (Pacific Biosciences) at Macrogen Inc. A total of 77.7 Gb continuous long reads (CLR) were assembled using the programme wtdbg ver. 2.5 with default parameters.²⁴ The programme Canu ver. 2.2²⁵ was also used for genome assembly, but the output was not adopted in later analyses because the output with wtdbg showed higher continuity (Table S2).

2.3. Hi-C data production and genome scaffolding

A Hi-C library was prepared using muscle tissue from the back near the base of the dorsal fin from the female (named sChiPha1) used for genome sequencing, as described in the iconHi-C protocol.²⁶ The library was digested with the restriction enzymes DpnII and HinfI and sequenced on a HiSeq X sequencing platform (Illumina Inc., CA, USA). This sequencing produced 263 million PE150 read pairs, amounting to a total of 79 Gb. The obtained Hi-C read pairs were processed with the programme Trim Galore! ver. 0.6.8 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) specifying the options ‘–phred33 –stringency 2 –quality 30 –length 25 –paired’ and aligned to the HiFi sequence contigs with the programme Juicer ver. 1.6. HiFi sequence contigs were scaffolded with 3d-dna (ver. 201008²⁷), specifying the options ‘-m haploid -i 1000 –editor-repeat-coverage 15 -r 2’ to be consistent with chromatin contact profiles. The continuity and completeness of the resulting genome assembly, designated sChiPha1.1, and the predicted protein-coding gene set were assessed using the webserver gVolante v2.0.0²⁸ in which the pipeline BUSCO ver. 5.1.2 was implemented,²⁹ using the ortholog set ‘vertebrate_odb10’ supplied with BUSCO. The completeness of the nucleotide sequences of the genome assemblies was assessed with compleasm (formerly called minibusco),¹⁰ which was designed to achieve higher accuracy through the use of miniprot.³⁰

2.4. Genome size estimation with k-mers

Raw short-read pairs were trimmed with SOAPnuke³¹ to remove low-quality bases and Illumina adaptor sequences. The processed reads were analysed by jellyfish v2.3.0 with a k-mer size of 31. The resultant.histo file was uploaded onto the GenomeScape webserver (http://genomescope.org/) to obtain genome size estimates.

2.5. Quantification of genomic rearrangements

To quantify cross-species karyotypic distance, we performed pairwise whole-genome alignment using minimap2 v2.26-r1175³² between the two holocephalans (the assembly built in this study for Ch. phantasma and GCF_018977255.1 for Ca. milii) and two elasmobranchs (GCA_030684315.1 for Stegostoma tigrinum and GCA_037974335.1 for Prionace glauca), with the parameters ‘-s 200 -z 200 -A 1 -B 1 -O 1,12 -E 4,1 -L 1 -m 50 -k 19 -c’ specifically tuned to detect macro-synteny across large evolutionary distances. The quantitative karyotypic distance was derived by meticulously parsing the resulting PAF file for structural discontinuities on a per-chromosome basis. This rearrangement score represents the minimum number of macro-syntenic breaks quantifying (1) changes in the target chromosome name between consecutive blocks on a continuous query chromosome, indicating a translocation breakpoint; (2) switches in the alignment strand orientation within the same query and target chromosome pair, signifying an inversion breakpoint; and (3) physical gaps exceeding a 100 Kb threshold between adjacent aligned segments on the same target chromosome, which were categorized as fission or fusion events.

2.6. Repeat identification

The obtained Ch. phantasma genome assembly was subjected to de novo repeat element identification with RepeatModeler ver. 2.0.4³³ with the option ‘-LTRStruct’, and the resultant repeat library was input into repeat masking using RepeatMasker ver. 4.1.5 with default parameters.³⁴

2.7. Transcriptome sequencing

Total RNA (0.5 to 2.7 μg) was extracted from the brain, kidney, ovary, gill, liver, heart, and muscle using an RNeasy Plus Universal Mini Kit (Qiagen, Hilden, Germany). RNA-seq was performed on total RNAs from each tissue. Libraries were prepared using the TruSeq Stranded mRNA Library Prep Kit (Illumina, CA, USA) and sequenced on NovaSeq 6000 to obtain PE reads with 150 cycles at Macrogen Inc.

2.8. Comparative analysis of the protein-coding gene landscape

The RNA-seq reads obtained as described above were mapped to the Ch. phantasma genome assembly using the programme STAR ver. 2.5.2b,³⁵ and the resultant bam file was input as evidence of transcription in protein-coding gene prediction to Braker ver. 2.1.5³⁶ incorporating all peptide sequences in the file, odb11_vertebrata_fasta, provided by OrthoDB.³⁷

Assessment of protein-coding gene structures and their genomic distributions was performed using the general feature format (GFF) file for Ch. phantasma output by Braker2 and GFF files retrieved from the NCBI Genomes database for the following species: Ca. milii (Accession ID: GCF_018977255.1), epaulette shark Hemiscyllium ocellatum (GCA_020745735.1), Japanese medaka (Oryzias latipes; GCF_002234675.1), chicken (Gallus gallus; GCA_016699485.1), and humans (Homo sapiens; GCF_009914755.1). From these GFF files, the lengths of individual introns were analysed to compute their length distributions for individual chromosome sequences for each species. For H. ocellatum, descriptions for intronic regions were amended by the programme agat_sp_add_introns.pl,³⁸ before the extraction of length distribution.

2.9. Orthology inference and cross-species comparisons of chromosomes

The above-mentioned protein-coding gene collection for the six species was modified to retain only peptide sequences with the longest open reading frames for individual genes, which were used as input for orthogroup construction with Orthofinder v2.5.5.³⁹ Homology of chromosomes between species was inferred using the locations of candidate orthologous genes classified by Orthofinder. Synteny comparison was visualized using nGenomeSyn⁴⁰ with single-copy orthologs detected by SonicParanoid ver. 2.0.8⁴¹ with the option ‘-m sensitive’.

2.10. Whole genome resequencing for sex chromosome identification

Genomic DNAs were extracted from nine males and nine females (Table S1) and used to obtain 20 to 30× paired-end short reads with PE150 on the DNB-seq sequencing platform at Macrogen Inc. (Table S1). The total volume of sequenced reads for females was 28.3 to 29.0 Gb and for males was 20.0 to 26.0 Gb. These reads, from which adapter and low-quality bases were removed (see above), were mapped to the genome assembly sChiPha1.1, with the mem option of bwa ver. 0.7.17-r1188,⁴² and the sequencing coverages from these BAM files were quantified with the bamcoverage option of deeptools ver. 3.5.4.⁴³ The average coverage in 10-Kb windows was compared between females and males.

For each individual, variants were called using GATK HaplotypeCaller⁴⁴ with the -ERC GVCF mode, and all gVCFs were merged with CombineGVCFs and jointly genotyped using GenotypeGVCFs. SNPs were extracted with SelectVariants and filtered with VariantFiltration using standard thresholds (QD < 2.0, FS > 60.0, MQ < 40.0, SOR > 3.0, MQRankSum < −12.5, ReadPosRankSum < −8.0).

Autosomal and sex-chromosomal SNP sets were then separated based on the identified X-chromosome sequence. For each individual, heterozygosity was calculated as the proportion of heterozygous SNPs (“0/1” or “1/0”) among all genotyped sites. Heterozygosity in males and females was summarized and compared using boxplots in R (v4.5.1).⁴⁵

2.11. Opsin gene search

To supplement protein-coding gene identification, the genome assembly sChiPha1.1 was input into the deep learning-based gene prediction programme Helixer ver. 0.3.2⁴⁶ using the model vertebrate_v0.3_m_0080.h5. The sources of the genome sequence data used to search for different opsin subtypes are listed in Table S3. Phylogenetic analyses to infer the orthology of identified opsin subtypes was performed as described previously.²⁰

3. Results

3.1. Genome sequencing and assembly

Continuous long-read (CLR) sequencing of the genome of Ch. phantasma produced a total read size of 77.7 Gb (76.9× of the resultant size of the obtained genome assembly) with an average length of 12.9 Kb. These reads were assembled into 1,029 contigs with an N50 length of 0.42 Mb; 29 sequences exceeded 10 Mb with a maximum sequence length of 19.9 Mb (Table S2). We separately prepared a Hi-C library and obtained 78 Gb paired-end read data (263 M PE150 pairs) from this library. The Hi-C data were used for Hi-C scaffolding into a chromosome-scale genome assembly (sChiPha1.1) containing 3,696 sequences with an N50 length of 42.7 Mb (Fig. 2) and a maximum sequence length of 155.4 Mb (Table 1). This assembly contained 1.01 Gb (Table 1), which closely resembled the genome size estimate based on k-mers, namely 1.04 Gb (Figure S1). In comparison with other genome assemblies of chimaeroid species, our genome assembly exhibited a higher completeness of genomic regions in chromosome-scale sequences (proportion of total bases in the scaffolds longer than 1 Mb: 96.0% versus 90.2% for Ca. milii), and conserved protein-coding genes shared as single-copy genes between vertebrates (completeness of single-copy orthologs: 96.8% versus 93.1% for Ca. milii) (Table 1). The modest score for duplicated orthologs (1.5%; Table 1) suggests the absence of recent lineage-specific whole-genome duplication.

Fig. 2.

Chromosomal organization. a) Chromatin contact map for Chimaera phantasma based on Hi-C data. This genome assembly consists of four chromosomes longer than 85 Mb and approximately 30 chromosomes shorter than 45 Mb. b) Chromosome length spectrum instructed by characterization of coding and non-coding genomic landscape. Repeat coverage denotes the percentages of interspersed repeats detected by RepeatMasker and simple tandem repeats detected by tantan (see Methods). The 50 longest sequences are shown.

Chromosomal organization of Ch. phantasma resembles that of Ca. milii, in which a relatively small number of chromosomes stand out in length: four chromosomes longer than 85 Mb and approximately 30 chromosomes shorter than 45 Mb (Fig. 2b). This pattern is the cytogenetic analyses of other chimaeroid species (H. colliei and Ch. monstrosa, respectively).¹² Although the chromosome numbers differ among species, their size composition is highly similar across the order. Despite our efforts, the karyotype of Ch. phantasma has not yet been obtained. The genomes of Ch. phantasma and Ca. milii share a gradual variation in the length of chromosomal sequences, which also resembles that of the chicken (Fig. 3a). Chromosome-level comparisons also revealed that shorter chromosomes exhibit higher GC-content, higher simple tandem repeat frequency, and higher gene density (Fig. 2b), as previously documented for avian chromosome organization (also see Introduction).^11,15

Fig. 3.

Comparative overview of genomic organization. a) Breakdown of whole genome assembly size into different chromosomal scaffolds. Chromosomal sequences are sorted by length, and scaffolds shorter than 0.2 Mb are excluded. b) Cross-species comparison of whole genome organization between Ch. phantasma and Ca. milii. Sequences of high similarity are shown with diagonal lines by the programme D-GENIES⁴⁷ with the “many repeats” mode.

We performed cross-species comparisons of chromosomal sequences within Holocephali, namely between Ch. phantasma and Ca. milii (Fig. 3b). This comparison, based on nucleotide sequence alignment, highlighted the unambiguous homology of Ch. phantasma chromosome 1 and Ca. milii chromosome 1, despite the intra-chromosomal breaks (Fig. 3b). Similarly, this comparison showed a one-to-one relationship for most chromosomes, except for chromosomes 5, 8, and 10 of Ca. milii, each of which corresponded to two Ch. phantasma chromosomes (Fig. 3b).

To perform an among-taxon comparison of the similarity of chromosomal organization taking intrachromosomal rearrangement, we quantified the breaks of nucleotide sequence alignments in the holocephalan pair (Ch. phantasma and Ca. milii) and an elasmobranch species pair (Stegostoma tigrinum and Prionace glauca) of similar divergence (see Methods). It detected 23,938 sites of rearrangement for the holocephalan pair, while the elasmobranch pair showed 117,530 sites.

3.2. Characterization of genomic compositions

A previous vertebrate-wide comparison revealed remarkable intragenomic heterogeneity in the base composition and distribution of protein-coding genes and repetitive elements.¹¹ In this study, we constructed a de novo repeat library and detected repetitive elements in the Ch. phantasma genome assembly and compared the repeat landscape with those of Ca. milii and H. ocellatum. This analysis revealed a similar profile for the genomes of Ch. phantasma and Ca. milii (Fig. 4). The total amounts of repeat elements (in both copy numbers and total bases) were much smaller in the two holocephalan species than in H. ocellatum and are dependent on genome size. The two holocephalan genomes contained more short interspersed nuclear elements (SINEs) than that of H. ocellatum, while the latter contained more long terminal repeats (LTRs) (Fig. 4).

Fig. 4.

Comparison of the repetitive element breakdown in two chimaera and one shark species. Proportions of genomic regions annotated to different repeat classes were obtained using RepeatModeler and RepeatMasker as described in the Methods. See Figure S2 for the divergence distribution of detected transposon sequences in these different species.

3.3. Cross-species comparison of protein-coding landscape

In the Ch. phantasma genome assembly, exons for protein-coding genes were inferred by incorporating evidence of transcription from RNA-seq data, as well as peptide sequences from other species (see Methods). In total, 24,724 protein-coding gene models were identified, resembling those of other vertebrates. To assess the effect of genome size variation on gene structure, we focused on intron length. A previous analysis of elasmobranch genomes showed that they have larger introns than other vertebrates, including Ca. milii (Hara et al.).²⁰ It is of great interest to identify which genomic segments are particularly responsible for this intron enlargement. In other shark species, large chromosomes tend to harbour large genes (including both exons and introns),¹¹ however, no analysis conducted to date has demonstrated a chromosome-level pattern with respect to intron size. Therefore, we compared the median length of introns in the predicted protein-coding genes between the different Ch. phantasma chromosomes. We observed a tendency for longer chromosomes to harbour protein-coding genes with longer introns (Pearson’s correlation coefficient, r = 0.96; Fig. 5). This pattern was shared among Ca. milii, H. ocellatum, and chicken genomes, whereas the human and medaka genomes did not exhibit this trend (r < 0.25; Fig. 5). This analysis suggests that the length variation of chromosomes can at least partly be linked to intron length variation, and its extent may contribute to the variation in whole genome sizes.

Fig. 5.

Intron size per chromosome. The vertical axis shows the median intron size and the horizontal axis shows the chromosome length. Sequences longer than 2 Mbp are regarded provisionally as chromosomes in the sChiPha1.1 assembly (see Discussion). Note that the scale for the vertical axis of H. ocellatum differs from that of the other panels. The Pearson's correlation coefficients in the panels were calculated by the cor function of R ver. 4.0.3.

The scatter plot in Fig. 5 shows some outliers. These chromosome outliers in chondrichthyan species tend to harbour a small number of protein-coding genes: in Ch. phantasma, scaffold 35 (2.09 Mb) contains only three predicted protein-coding genes; in Ca. milii in which even sequences shorter than 1 Mb are labelled as ‘chromosomes’ (Nakatani et al.),⁸ chromosomes 34 (933 Kb) and 35 (792 Kb) contain only four and five genes, respectively. The paucity of genes on these ‘chromosomes’ prevents us from judging whether these sequences have unusual genomic characteristics or whether this is caused by the small sample size. It is also possible that these are fragments of longer chromosomes that were not properly assembled or were scaffolded.

3.4. Identification of sex chromosome sequences

Holocephalan sex chromosomes have not been identified in cytogenetic or genomic analyses. This prompted us to search for candidate sex chromosomes in the Ch. phantasma genome assembly. For this purpose, we performed whole genome resequencing of eight male and female Ch. phantasma (that did not include the individual used for long-read sequencing), and quantified the sequencing coverages of individual genome scaffolds in the assembly sChiPha1.1 that was produced in the present study. A comparison between males and females revealed a ratio of approximately 0.5 for Ch. phantasma scaffold 41 (Fig. 6), suggesting that this scaffold was sequenced twice in females compared to other genomic regions. Assuming that this species has an XY sex chromosome system (male heterogamety), scaffold 41, which is approximately 500 Kb in length, may represent a fragment of the X chromosome. Consistent with this inference, SNP-based heterozygosity on scaffold 41 in males (Scaf. 41-M, Fig. 7) differed markedly from that on the autosomes, with heterozygosity being nearly zero, as expected for a hemizygous region. Although we observed other genomic regions with deviating ratios, scaffold 41 was the only one with a consistent deviated pattern in its entire sequence stretch.

Fig. 6.

Male-female comparison of sequencing depth for identifying sex-related genomic regions. The upper panel shows all the scaffolds including chromosome-scale sequences, and the lower panel shows its subset, scaffolds 35 to 45.

Fig. 7.

Comparison of heterozygosity between males and females across genomiwide (autosome) and the putative X chromosome, namely scaffold 41.

The Ch. phantasma scaffold 41 was predicted to harbour 61 protein-coding genes, which did not include any orthologs of the genes identified as master sex determination genes in other species.^48–51 Orthologs of the 61 genes were most frequently located on chromosome 1 of the human genome and chromosome 25 of the chicken genome (Fig. 8a). Conversely, most orthologs of genes on the human X and Y chromosomes were located on scaffolds 4 or 13 of the Ch. phantasma genome, and most of those on chicken Z and W chromosomes were found on scaffold 1 of the Ch. phantasma genome (Fig. 8a). This comparison with human and chicken genomes showed the distinct autosomal origin of Ch. phantasma X chromosome compared to those non-chondrichthyan lineages with the established sex chromosomes.

Fig. 8.

Chromosomal homology of C. phantasma inferred by one-to-one ortholog localization. a) Genome-wide comparison with human (GRCh38.p14) and chicken (GRCg7b). b) Genome-wide comparison among chondrichthyans. For C. phantasma, the 50 largest scaffolds (including the putative X chromosome, namely scaffold 41) are shown, while all the sequences of H. ocellatum (GCF_020745735.1) and S. tigrinum (GCA_022316705.1) registered as chromosomes in NCBI are shown. The one-to-one orthologies of protein-coding genes on the C. phantasma X chromosome sequence are shown in red, while those on the X chromosomes of the other species are shown in light green. See Figure S4 for the version with a colouring for all chromosomes.

Many of the orthologs of the genes predicted on Ch. phantasma scaffold 41 were located on chromosome 50 of the H. ocellatum genome and chromosome 47 of the S. tigrinum genome (Fig. 8b); these two chromosomes are not sex chromosomes in these elasmobranch species. The S. tigrinum X chromosome has been shown to be homologous to a part of human chromosome 12 and to chicken chromosome 34.¹¹ This putative elasmobranch X chromosome harbours Dhh, Gdf11, and a cluster of Wnt1, −6b, and −10b genes in its X-specific region, as well as HoxC genes in its pseudoautosomal region. Our search for a region in the Ch. phantasma genome that is homologous to this S. tigrinum X chromosome as well as H. ocellatum genome consistently identified Ch. phantasma scaffold 29 (6.1 Mb) (Fig. 8b). This scaffold harbours orthologs of Dhh, Gdf11, Wnt1, −6b, −10b, and HoxC. The Ca. milii genome possesses a highly constrained Hox C cluster as in many other vertebrates,^19,52 and we found that the Ch. phantasma genome also harbours a conserved Hox C cluster with an identical set of gene repertoires to Ca. milii (HoxC1, -C3, -C4, -C6, -C8, -C9, -C10, -C11, -C12, and -C13).

3.5. Opsin gene repertoires

Previous studies characterized the repertoires of opsin genes in some chondrichthyan species.^2,3,53 To infer the reliance of Ch. phantasma, a deep-sea dweller, on photoreception, we utilized its genome assembly. Gene models from our primary gene prediction included only three opsin genes (rhodopsin, pinopsin, and opsin 3), which were supplemented by secondary gene prediction using a deep-learning-based programme and manual curation of putative open reading frames (Table S4). In these additional approaches, the former source contained VA (vertebrate ancient) opsin, TMT (teleost multiple tissue) 1, TMT2, opsin 4m, opsin 6a, opsin 7b/c and RRH (retinal pigment epithelium-derived rhodopsin homolog), while the latter harboured opsin 4x, opsin 5L2, and RGR (retinal G protein coupled receptor) (Fig. 9a), based on the previously proposed nomenclature.² Remarkably, the orthologs of Rh2, LWS (long wavelength-sensitive), parapinopsin-like, TMT3, opsin 6b, and opsin 5 were not identified as intact ORFs in the Ch. phantasma genome (Fig. 9a), while their orthologs have persisted in the Ca. milii genome.^2,3,57 For Rh2, TMT3, and Opn6b, the adjacent linkage of the two flanking genes was maintained in the holocephalan species, and the intergenic stretch contained no undetermined regions (red lines in Fig. 9b). This type of evidence could not be obtained for LWS and parapinopsin-like genes whose Ca. milii orthologs were found on an end of scaffold sequences shorter than 100 Kb. For each of Opn5m, TMT1, TMT2, we identified a region in the Ch. phantasma genome in which translated peptide sequences show substantial similarity to the Ca. milii ortholog but contained multiple termination codons, suggesting pseudogenization of these Ch. phantasma orthologs (Fig. 9a).

Fig. 9.

Genomic characterization of Ch. phantasma opsin gene repertoires. a) Opsin gene repertoires. The numbers in the boxes indicate the numbers of paralogs generated in its unique lineage. Presence and absence of orthologs were determined based on the survey in both predicted peptide sequences and genomic nucleotide sequences of individual species. Only those confirmed with putative full-length ORFs supported by cross-species sequence homology are shown with coloured boxes as possessing an ortholog, and possible Ch. phantasma pseudogenes with disrupted ORFs are shown as open boxes. Colouring of Rh2, SWS1, SWS2, and LWS in this figure is based on typical maximal absorbances of pigments encoded by them. See¹¹ for details of possible pseudogenes excluded from this presentation and categorization of human genes (asterisk). b) Opsin gene loss in the Ch. phantasma genome. Absence of the opsin ORF was confirmed in the stretch between the two genes flanking the opsin subtype gene of question retained in the Ca. milii genome. Only the cases confirmed with sufficiently long scaffolds are depicted. c) Key amino acid residues potentially contributing to spectral tuning of rhodopsins. The phylogenetic relationships among the species involved are depicted based on existing literature,⁵⁴ with previously documented spectral tuning sites. Habitat depths are based on existing literature.^5,55,56

Species living in dim-light environments tend to harbour amino acid substitutions in the visual pigment rhodopsin (RHO) that alter the absorption spectrum to effectively sense light near the wavelength of 480 nm, conventionally called the ‘blue shift’.⁵⁸ We assessed the deduced amino acid sequence of the rhodopsin ortholog of Ch. phantasma, which was shown to exist as a single ortholog in its whole genome (Fig. 9a). Comparison with other cartilaginous fishes revealed a substitution from alanine to serine at residue 292 of the bovine RHO sequence (Fig. 9c). This substitution was documented to be shared with two deep-sea dwellers, namely, the cloudy catshark Scyliorhinus torazame and blackmouth catshark Galeus melastomus, for which a blue shift was previously suggested.^2,20,59

4. Discussion

4.1. Trends in holocephalan karyotypes based on DNA sequences

In this study, we performed whole genome sequencing for Ch. phantasma to obtain a chromosome-level assembly, as the second species in Holocephali following Ca. milii, and the first in the family Chimaeridae. This genome assembly provides a valuable resource for validating the evolutionary processes of early vertebrates inferred in past studies using the Ca. milii genome alone.⁷

The subclass Holocephali is species-poor, with a maximum of 60 species; however, it occupies a unique phylogenetic position with approximately 400 million years of divergence from Elasmobranchii, the other extant chondrichthyan lineage. In concordance with the results of cytogenetic studies on other holocephalan species,¹² we found that the Ch. phantasma genome assembly contained only four chromosomal sequences longer than 85 Mb, while the fifth longest sequence was shorter than 45 Mb (Fig. 2b). A relatively large gap in size between the large and small chromosomes may be a characteristic of holocephalan species, given that the genome assembly of Ca. milii also showed a similar pattern of chromosome size variation.⁸ However, this interpretation needs to be supported by information from other species, especially those in Rhinochimaeridae for which no genome assembly (even on a non-chromosome level) is available as of May 2025.

4.2. Genomic organization within chondrichthyans

Compared with other vertebrates, chondrichthyan genomes have a wider distribution of chromosome lengths, with dozens of chromosomes shorter than 30 Mb (Fig. 4). A high number of short chromosomes in karyotypes has also been observed in most bird and reptile species, and these small chromosomes are termed ‘microchromosomes.’ This pattern of karyotype variation has not been observed in teleost fish or mammalian species (Fig. 3a). Nevertheless, there are remarkable differences among chondrichthyan species. The analysed holocephalan species have only a few chromosomes longer than 50 Mb, whereas elasmobranchs have dozens.^11,20,60 Our between-taxon comparison detected fewer chromosomal rearrangements in holocephalan whole genome alignment than in an elasmobranch counterpart. Even though the genome size difference between the species pairs is considered, the difference remarkably suggests a more stable intra-chromosomal genomic organization in holocephalans.

The range of genome sizes differs between Holocephali and Elasmobranchii. Species in the latter taxon often have genome sizes larger than 3 Gb, with a current maximum of approximately 15 Gb,⁶¹ whereas the genomes of holocephalans are reported to be smaller.¹⁶ This is exemplified by the relatively small size (1.01 Gb) of the Ch. phantasma genome assembly produced in the present study and that of Ca. milii (Table 1). One caveat to this conclusion is that the number of holocephalan genomes investigated remains very small. Feulgen staining analysis of the genome size of Ca. milii provided an estimate of 1.9 Gb.⁶² This was clearly larger than the genome assembly size (Table 1). Such gaps, sometimes exceeding 1.0 Gb, were also identified in earlier studies on elasmobranchs (reviewed in⁶¹). One possible reason may be the highly repetitive nature of the genome.

4.3. Chondrichthyan sex determination studies

The present study provides the first identification of DNA sequences of a candidate X chromosome in a holocephalan species (Figs. 6–8). This sequence consistently showed an approximately doubled sequence depth in females and nearly complete lack of SNP heterozygosity in males, but its length of 500 Kb is short for an entire chromosome in comparison with typical vertebrate chromosomes. In the chicken, which has numerous microchromosomes, all chromosomes are larger than 2 Mb in the latest telomere-to-telomere genome assembly.⁶³ It is likely that the sequencing and assembly of the Ch. phantasma X chromosome sequencing suffered from the typical difficulty of high GC-content and high repetitive element density,⁶⁴ which may have resulted in its small length.

The putative X chromosome sequence of Ch. phantasma contains 61 predicted protein-coding genes, but none were orthologous to the master sex determination genes namely, Sox3 (eg, Sry), Hsd17b1 or Dmrt1, as well as a member of TGF-β superfamily and its receptors, such as Amh, Amhr2, Bmpr1b, Gsdf, and Gdf6, previously reported in other vertebrates including mammals, birds, and teleost fishes.^49–51 Within the Chondrichthyes, X chromosome sequences were reported for S. tigrinum, R. typus, H. ocellatum, and C. plagiosum, and those chromosomes were shown to be homologous to each other.^11,22,65,66 Some genome assemblies of elasmobranchs in public sequence databases include chromosomes labelled as X or Y, but these identities have yet to be substantiated.^60,64–66 In particular, sequence comparisons²² showed the homology of H. ocellatum X chromosome to the X chromosome of S. tigrinum in which the X chromosome identification was based on firm male-female comparisons with whole genome sequencing and quantitative PCR experiments.¹¹ In the precent study, our comparison of ortholog locations showed non-homology of the Ch. phantasma X chromosome sequence (scaffold 41) to the putative H. ocellatum X chromosome (Fig. 8b). These species both display male heterogamety, but genomic comparisons show discordance in autosomal origins of those X chromosomes.⁶⁷ Investigations on other chondrichthyan lineages including batoids (rays and skates), as well as on more chimaeroids, are expected to elucidate which represents the ancestral pattern of sex chromosome organization and how this discrepancy arose. One possibility is that turnover of the sex chromosome or extensive rearrangements of the X chromosome likely occurred in the lineage leading to Ch. phantasma, after the split between the holocephalan and elasmobranch lineages. The present study used a female of Ch. phantasma that should lack the Y chromosome for assembling a reference genome, and further investigation are required to identify Y chromosome sequences that will provide clues for understanding the sex determination mechanism in chondrichthyans.

4.4. Lineage-specific modification of opsin gene repertoires

Our genome-wide survey revealed a variable pattern of opsin gene retention between the two holocephalan species—the Ch. phanstama genome retains fewer opsin genes, including the rhodopsin whose amino acid sequence harboured a typical substitution observed for deep-sea dwellers. The habitat depth of Ca. milli and Ch. phantasma overlap, with Ca. milli recorded at 0 to 227 m and Ch. phantasma at 20 to 962 m.⁵ However, Ca. milli is most commonly sampled at depths shallower than 100 m.⁶⁸ In contrast, Ch. phantasma has been recorded at approximately 500 m in Japanese waters.⁶⁹ Based on the author’s field surveys and locality data of specimens registered in Japanese museums, Ch. phantasma primarily occurs at depths of 200 to 500 m, with records from shallower than 100 m being extremely rare. These observations suggest that Ca. milli mostly inhabits bright, shallow water zone, whereas Ch. phantasma is adapted to dimmer and bluer, mesopelagic zone.⁷⁰ The difference in habit ranges may account for the difference of the rhodopsin sequences and the overall opsin gene repertoires between the two holocephalan species.

Supplementary material

Supplementary data are available at DNARES online.

Funding

This work was supported by NIG-JOINT (3B2022) to S. Hirase, “Strategic Research Projects” grant from ROIS (Research Organization of Information and Systems) to S. Kuraku, Grants-in-Aid for Scientific Research (22H00377) to K. Kikuchi and Grant-in-Aid from JSPS Fellows (DC1 202022659) to A. Teramura.

Data availability

The genome assembly of Ch. phantasma and the gene models associated to it are deposited at https://figshare.com/s/0db52bd7af0a3961bf8e  https://figshare.com/s/a8e1af30d50aa5e3dc91, and https://figshare.com/ndownloader/files/54445589, respectively. These data, as well as the associated raw sequence reads, have been deposited in NCBI under BioProject PRJNA1075468.