Hemoglobin-Gene Cluster Deletions in Antarctic White-Blooded Icefishes Facilitated by Transposable Elements

Abstract

Vertebrates transport oxygen throughout the body bound to hemoglobin packed in red blood cells. Antarctic icefishes are evolutionary oddities with transparent, white blood lacking hemoglobin. Previous work revealed that the icefish ancestor lost all hemoglobin genes, except for one exon of one alpha-globin gene. This peculiar condition has puzzled biologists for decades and is widely considered maladaptive because it is associated with numerous physiological and anatomical compensations. While we know that hemoglobin genes evolved episodically during cold adaptation in Antarctic notothenioid ancestors and that red-blooded sister species to icefishes have seemingly altered hemoglobin and lack multiple hemoglobin isoforms (hemoglobin multiplicity), we still do not know how the two hemoglobin gene clusters, the LA and MN clusters, shared by all other teleost fishes, were lost in the icefish ancestor. Here, we analyzed the two hemoglobin cluster regions in ten red-blooded notothenioid species and their orthologous region in eight icefishes and identified traces of molecular mechanisms that could have independently deleted each cluster. Adjacent to the LA cluster, an Antarctic notothenioid-specific massive amplification of tRNA genes associated with transposable elements (TEs) likely deleted the hemoglobin cluster when tRNAs transposed into the second intron of the second alpha-globin gene. Independently, extreme species-specific gene turnover characterized the evolution of the MN cluster in all notothenioids. Conserved TEs precisely located at the boundaries of the deleted MN hemoglobin gene cluster incriminate them in the cluster's deletion. Together, these results provide a deeper understanding of the evolution of the iconic icefishes and their unique phenotype.

Significance.

Antarctic white-blooded icefishes are the only vertebrates living without hemoglobin. While we know that hemoglobin genes were lost in the icefish ancestor after its divergence from dragonfishes, we still do not know the underlying mechanisms that led to this unique hemoglobin-less condition. Here, we analyzed the two genomic regions containing hemoglobin clusters in red-blooded notothenioid species and revealed that the orthologous genomic regions in icefishes bear traces of molecular mechanisms that could have deleted them. Genomic analyses further demonstrated that in icefishes, the two clusters were lost independently from one another by different genomic mechanisms, although both mechanisms incriminate transposable and repeat-like elements. These analyses provide the first evidence-based mechanistic models for how hemoglobin genes were lost in the icefish ancestor and thus bring us closer to understanding the deep evolution of this iconic clade.

Introduction

Hemoglobin is a tetrameric protein composed of two alpha- and two beta- or beta-like subunits that transports oxygen in red blood cells in nearly all vertebrates, the sole exception being the Antarctic icefishes (Channichthyidae), which lost all coding sequences of hemoglobin genes, retaining only the third exon of a single alpha-globin gene (Near et al. 2006; Storz 2018; Kim et al. 2019). Because of the absence of hemoglobin, the blood of icefishes is translucent white (see Sidell and O’Brien (2006) and Beck et al. (2022) for images). Icefishes transport oxygen only by dissolution in the plasma, which represents only ∼10% of the oxygen carried by the blood of their red-blooded relatives expressing hemoglobin (Ruud 1954; Holeton 1970). While the loss of hemoglobin genes happened in a common ancestor of all icefishes after its divergence from Antarctic dragonfishes about 5.3 million years ago (MYA) (Eastman 1993; Near et al. 2006; Bista et al. 2023), we do not know the genetic mechanisms that caused this deletion. It also remains unknown whether these deletions would have been adaptive, neutral, or maladaptive (Sidell and O’Brien 2006; Daane et al. 2020). This profound state of anemia, however, required major cellular and physiological compensations that included a dramatic increase in mitochondrial densities and the remodeling of the vascular system with an enlarged heart and extended vasculature compared to red-blooded relatives of similar size (e.g. Sidell and O’Brien 2006; Wujcik et al. 2007; Garofalo et al. 2009; Damsgaard et al. 2019; Beck et al. 2022).

Hemoglobin gene clusters in sharks, ray-fin fishes, and lobe-fin fishes like coelacanth and amphibians, each contain genes for both alpha and beta hemoglobin chains, but in amniotes, including reptiles, birds, and mammals, alpha and beta chain genes are in different gene clusters due to a translocation event after the divergence of amniotes from amphibians (e.g. Cooper et al. 2006; Fuchs et al. 2006; Opazo et al. 2008; Patel et al. 2008; Hardison 2012; Schwarze and Burmester 2013; Opazo 2015a,b; Philipsen and Hardison 2018). In teleost fishes, hemoglobin genes are organized in two genomic clusters: the “LA” and “MN” clusters that originated from the duplication of the ancestral vertebrate hemoglobin cluster with the Teleost Genome Duplication (Postlethwait et al. 1998; Opazo et al. 2013). These two clusters were named based on the two genes flanking one side of each cluster: lcmt1 and aqp8 for the LA cluster, and mpg and nprl3 for the MN cluster (Hardison 2008; Opazo et al. 2013). Within each teleost cluster, hemoglobin genes are usually organized in alternating alpha- and beta-globin genes that are on opposite DNA strands (Opazo et al. 2013; Hotaling et al. 2023; Omelchenko et al. 2024). The repertoire of hemoglobin genes across teleost species is highly variable, especially in the MN cluster in which gene number can range from zero in Ishikawa noodlefish to 26 in Northern Pike (Opazo et al. 2013; Bista et al. 2023; Desvignes et al. 2023; Hotaling et al. 2023; Omelchenko et al. 2024; Li et al. 2025). Among Antarctic notothenioids (cryonotothenioids), an abundance of transposable elements (TEs) in both hemoglobin clusters was proposed to have facilitated the tandem duplication of gene pairs by nonhomologous recombination (Bista et al. 2023).

We recently described the evolution of both LA and MN hemoglobin clusters in Antarctic notothenioids (Bista et al. 2023) and revealed that notothenioid hemoglobin genes were subject to diversifying selection during cold adaptation and again during the diversification of lineages closely related to icefishes (Desvignes et al. 2023). In addition, while teleost fish generally express multiple hemoglobin proteins with differing physiochemical properties (i.e. hemoglobin multiplicity) through the expression of diverse combinations of alpha- and beta-globin paralogs, adult Antarctic notothenioids display reduced hemoglobin multiplicity (e.g. di Prisco et al. 2007; Verde et al. 2012; Giordano et al. 2015). Diminished hemoglobin multiplicity was caused by the reduction or loss of hemoglobin gene expression from the MN cluster, and the predominant expression of only one alpha- and one beta-globin gene from the LA cluster (Desvignes et al. 2023). These observations suggested that the icefish ancestor was also likely relying on the expression of a single pair of hemoglobin genes (Desvignes et al. 2023). In that previous study, however, gene and protein sequence analyses were insufficient to resolve the mechanisms by which both hemoglobin gene clusters were lost in icefishes.

Here, analyses using 18 highly contiguous notothenioid genome assemblies (Table S1) revealed that the hemoglobin cluster genomic regions in icefishes bear traces of putative molecular mechanisms that could have deleted them and demonstrated that in icefishes, the LA and MN clusters were lost independently from one another by different genomic mechanisms, although both mechanisms relate to the expansion of transposable and repeat-like elements.

Results and Discussion

Dramatic Expansion of the LA Cluster Region in Cryonotothenioids

As previously described (Bista et al. 2023; Desvignes et al. 2023), the organization and content of the LA cluster remained well conserved during notothenioid evolution (Fig. 1a). The only copy-number variation involved the beta-globin gene hbbla1 (hereafter β-la1) that was tandemly duplicated in the common ancestor of the Patagonian Blennie Eleginops maclovinus and cryonotothenioids after its divergence from bovichtids (Bista et al. 2023; Desvignes et al. 2023). In most Antarctic notothenioid lineages studied, the hbbla1.2 duplicate (β-la1.2) was repeatedly pseudogenized or lost, while the hbbla1.1 duplicate (β-la1.1) significantly diverged in sequence but was preserved in all nonicefish species investigated (Bista et al. 2023; Desvignes et al. 2023).

Fig. 1.

LA cluster region organization. a) Phylogenetic relationships and schematic representation of the genomic organization of alpha-globin genes (hbala) and beta-globin genes (hbbla) in the LA cluster and neighboring genes. Arrows for the entire cluster represent orientations in each assembly. b) Dot plots of self-alignments for the LA region and flanking genes. Green and red represent forward and reverse alignments, respectively. Grey, blue, pink, and yellow shadings highlight the flanking genes, the LA cluster, chrn-like genes, and tRNA genes, respectively. c) Bar chart of the number of tRNA genes in the LA Region in the studied species. d) Phylogenetic regression of the number of tRNA genes in the LA region by the LA region size. White, green, blue, and cyan dots label outgroup species, temperate notothenioid species, red-blooded cryonotothenioids, and icefishes, respectively. Species' dots are connected following the phylogenetic relationships shown in a), and small grey dots mark internal tree nodes. The dashed line represents the regression, and its equation and support appear in the lower right corner. e) Dot plots of species pair alignments for the portion of the LA region containing the hemoglobin genes and the flanking gene aqp8.2. Horizontal and vertical axes are drawn to scale, according to the sequence length (in base pairs) of the region in each respective species.

In contrast to a decrease in the number of hemoglobin protein-coding genes, a dramatic increase in the size of the genomic region of the LA cluster between its two flanking genes, rhbdf1b upstream and aqp8.2 downstream of the cluster, occurred with the diversification of cryonotothenioids (Table 1, Table S2). Note that we define the “LA region” as the segment between its two flanking genes rhbdf1b and aqp8.2, while the “LA cluster” is the nested region between the start codon of hbala1 (α-la1) and the stop codon of hbala2 (α-la2). In three non-Antarctic notothenioids, as well as in all ten other Perciformes previously investigated, the LA region spans on average 49 ± 21 kb, and ranges from 13 kb in the Lumpfish Cyclopterus lumpus to 97 kb in the snailfish Pseudoliparis sp. from the Yap trench (Hotaling et al. 2023). In red-blooded cryonotothenioids, however, the orthologous LA region is seven times larger, spanning on average 385 ± 131 kb and ranging from 231 kb in the Patagonian Toothfish Dissostichus eleginoides to 572 kb in the Antarctic Toothfish Dissostichus mawsoni (Table 1). Thus, while the LA cluster itself occupies 24 ± 14% of the LA region in 13 noncryonotothenioid perciform species, the LA cluster represents only 5 ± 1% of the LA region in four red-blooded cryonotothenioids (Table 1). In four icefishes, although only the 129 bp-long third exon of the α-la2 gene remains to represent the LA cluster, the LA region spans on average 208 ± 53 kb and ranges from 159 kb in the Pike Icefish Champsocephalus esox to 263 kb in the Mackerel Icefish Champsocephalus gunnari, thus still several times longer than in non-Antarctic cryonotothenioids (Table 1).

Table 1.

Size and content of the LA region in studied species

Species	Group	LA region length (kb)	LA cluster length (kb)	LA cluster size (% LA region)	# chrn-like pseudogenes	# tRNA genes
Perca flavescens	OUT	73	14	19	0	22
Ophthalmolycus amberensis	OUT	64	7	11	0	40
Cottoperca trigloides	TEMP	38	8	22	0	47
Bovichtus variegatus a	TEMP	64	6	10	0	3
Eleginops maclovinus a	TEMP	44	10	23	0	29
Dissostichus mawsoni a	CRYO	572	21	4	4	354
Dissostichus eleginoides	CRYO	231	16	7	2	124
Trematomus bernacchii a	CRYO	430	16	4	6	209
Notothenia rossii	CRYO	298	14	5	3	317
Harpagifer antarcticus	CRYO	NA	13	NA	NA	76
Pogonophryne albipinna a	CRYO	395	15	4	1	175
Gymnodraco acuticeps	CRYO	NA	17	NA	NA	26
Champsocephalus gunnari a	ICE	263	0	0	2	313
Champsocephalus esox a	ICE	159	0	0	1	108
Pseudochaenichthys georgianus a	ICE	167	0	0	1	103
Chaenocephalus aceratus a	ICE	243	0	0	0	250

OUT, outgroup; TEMP, early diverging temperate notothenioids; CRYO, red-blooded cryonotothenioids; ICE, Icefishes.

^aDenotes that the LA region contains one or more gaps in that genome assembly.

To explore the content of the LA region, which does not appear to contain functional protein-coding genes, we produced self-alignments of the LA region (i.e. the alignment of a species' LA region against itself) in 12 notothenioid and 2 non-notothenioid species for which the region is assembled into a single scaffold, although the assembled region of 9 of them still contains gaps. These self-alignments revealed extensive blocks of short, repeated sequences in both temperate and Antarctic species (Fig. 1b, Fig. S1). These short repetitive sequences were also found in the LA region of the non-Antarctic notothenioids Channel Bull Blenny Cottoperca trigloides (Fig. 1b1) and E. maclovinus (Fig. S1), but in far lower proportions than in red-blooded cryonotothenioids (Fig. 1b2, Fig. S1). Additionally, patterns of repeated sequences multiple kb long were also apparent in red-blooded cryonotothenioid species (Fig. 1b2, Fig. S1). BLAST alignments of these longer repeated sequences against the nucleotide database in NCBI identified them as multiple chrn-like (cholinergic receptor, nicotinic-like). All these chrn-like genes, however, appear to be pseudogenes because each duplicate misses one or multiple exons found in functional chrn genes. A search for paralogous genes in temperate and Antarctic notothenioid genome assemblies revealed that an additional repetitive cluster of chrn-like genes is present in all studied notothenioid species, including icefishes, and is located on the same chromosome as the LA cluster, however, approximately 11 ± 3 Mb away from the LA region (Table S2). In temperate notothenioids, no chrn-like genes or pseudogenes were found in the LA region, while in red-blooded cryonotothenioids, their number ranged from one chrn-like pseudogene in the Whitefin Plunderfish Pogonophryne albipinna to six chrn-like pseudogenes in the Emerald Notothen Trematomus bernacchii (Table 1). In icefishes, three of the four studied species possess one pseudogenized chrn-like gene (Fig. 1b3), but none were found in the Blackfin Icefish Chaenocephalus aceratus (Table 1). Thus, we conclude that the chrn-like gene cluster was duplicated and translocated to the LA region in an ancestor of cryonotothenioids after it diverged from the lineage of the temperate species E. maclovinus. The large variation in chrn-like copy-number across cryonotothenioids and the relative short length of the chrn-like genes compared to the total length of the LA region in all cryonotothenioids, however, suggest that this translocation was not the main driver of LA region length expansion.

tRNA-Driven Expansion of the LA Cluster Genomic Region in Cryonotothenioids

Gene annotations in the NCBI genome browser predicted many tRNA genes in the LA region of C. trigloides. We therefore extracted all LA regions that had been assembled into a single contiguous scaffold, and de novo annotated tRNA genes using tRNAscan-SE (Chan et al. 2021) (supplementary material S1). This reannotation of tRNA genes revealed that the highly abundant short, repeated sequences in the LA region correspond to tRNA genes. In C. trigloides, 47 tRNAs were predicted in the LA region, including 2 tRNAs located in the intergenic region between α-la1 and β-la1 and 1 in the first intron of α-la2 (Fig. 1b1 and c). The presence of numerous tRNAs in the LA region appears to be shared by all studied percomorph species, including the Yellow Perch Perca flavescens and the Antarctic eelpout Ophthalmolycus amberensis, with 22 and 40 tRNAs in the LA region, respectively (Fig. 1c). The number of tRNAs, however, dramatically increased in cryonotothenioids and ranges from 124 tRNAs in D. eleginoides to 354 in D. mawsoni for red-blooded cryonotothenioids (Fig. 1b2 and c), and from 103 in P. georgianus to 313 in C. gunnari for icefishes (Fig. 1b3 and c). Furthermore, the number of tRNAs and the size of the LA region were significantly correlated (P < 0.001, phylogenetic linear regression), although their number in cryonotothenioids is somewhat erratic compared to noncryonotothenioid species (Fig. 1d). These results reveal that the tRNA content of the LA region in cryonotothenioids is highly dynamic, even between sister species such as the Dissostichus and Champsocephalus species pairs (Fig. 1c). Additionally, these findings highlight that the LA region underwent lineage- and species-specific shuffling leading to major tRNA copy-number variations that appear to be the main driver of the changes in LA region size.

Among tRNA types, the LA region of cryonotothenioids is marked by the extreme multiplication of Isoleucine (Ile) and Threonine (Thr) tRNAs, and specifically in icefishes, of Arginine (Arg) tRNA (Fig. 1c). To identify whether this increase in tRNA number in the LA region of cryonotothenioids represents a genome-wide trend or is unique to the LA region, we annotated tRNAs genome-wide in the assemblies of 11 notothenioid and two non-notothenioid species (supplementary material S2). Results revealed that tRNA genes also dramatically multiplied genome-wide; however, the genome-wide increase was not specifically marked by an increase in Isoleucine or Threonine tRNAs as in the LA region (Fig. S2a). Nonetheless, the number of tRNA genes in the LA region was significantly correlated with the number of tRNAs genome-wide in the assemblies (P = 0.018, phylogenetic linear regression) (Fig. S2b). However, if the increase in tRNA number in the LA region is the result of a general genome-wide increase in tRNAs, the proportion of genome-wide tRNAs present in the LA region would remain constant across species and would not be correlated with the number of tRNAs present in the LA region. Instead, the number of tRNAs in the LA region was significantly positively correlated with the proportion of genome-wide tRNAs present in the LA region (P = 0.0017, phylogenetic linear regression) (Fig. S2c), meaning that the number of tRNAs in the LA region increased proportionally more than in the rest of the genome assembly. This finding thus suggests that the LA region is a “hotspot” of tRNA gene increase in cryonotothenioid genomes. Furthermore, if the LA region were the only or the major “hotspot” of tRNA multiplication in the genome, the proportion of genome-wide tRNAs present in the LA region would be positively correlated with the number of genome-wide tRNAs. However, the proportion of genome-wide tRNAs present in the LA region was not correlated with the total number of tRNAs in the genome (P = 0.97, phylogenetic linear regression) (Fig. S2d). This lack of correlation suggests that the genome-wide increase in tRNAs was not solely driven by an increase in tRNAs in the LA region, thus other hotspots of tRNA multiplication are present in the notothenioid genomes.

These findings raised the hypothesis that the multiplication of tRNA genes in the LA region might be related to the deletion of icefish LA cluster hemoglobin genes. To analyze this possibility, we carefully inspected the part of the LA region flanking the hemoglobin genes using dot plots between species pairs. Alignment of the LA region of the temperate C. trigloides with that of the Antarctic dragonfish Gymnodraco acuticeps revealed that tRNAs are present a few kb away from α-la1 in both of these red-blooded species, as well as in all other studied red-blooded notothenioids (for example, the 44 tRNAs in Fig. 1e1 to 2; Figs. S3 and S4). In contrast to all red-blooded species studied, a small cluster of a few tRNAs was consistently located much closer, only hundreds of base pairs, away from the remaining α-la2 exon in all icefish species studied (for example, the 6 tRNAs found only 442 base pairs away from the α-la2 exon in Fig. 1e3, Figs. S5 and S6). One of these tRNAs, a tRNA-Ile, had previously been reported in the vicinity of exon-3 of α-la2 in most icefish species (Near et al. 2006). The consistent proximity of these tRNAs to the hemoglobin pseudogene among icefishes suggests that tRNA insertions or amplification in the vicinity of exon-3 of α-la2 occurred in an ancestor of icefishes after the divergence of icefishes from dragonfishes about 5.3 MYA (Bista et al. 2023). This discovery suggests that tRNA amplifications or insertions may be related, not only in timing but also potentially in mechanism, to the loss of hemoglobin genes in the icefish ancestor.

TE-Driven Deletion of the LA Cluster Hemoglobin Genes in the Icefish Ancestor

We then examined the mechanisms by which tRNA genes multiplied in the LA region and translocated next to the exon-3 of α-la2 in the icefish ancestor. Because TEs are often associated with genome rearrangements caused by transposition or recombination events between homologous regions and can lead to copy-number variations of genes neighboring the TEs (e.g. Bourque et al. 2018; Wells and Feschotte 2020; Balachandran et al. 2022), we hypothesized that DNA sequence repeats and TEs could have facilitated these tRNA gene multiplications; hence, we first analyzed the repeat and TE content of the LA region compared to genome-wide TE content and distinguished TEs by main classes (e.g. LINE, SINE, DNA, and LTR; supplementary material S3).

Consistent with the dramatic size increase of the LA region in red-blooded cryonotothenioids and in icefishes, we observed a large increase in young TEs and repeats (i.e. with low Kimura Divergence) in the LA region of red-blooded cryonotothenioids and icefishes compared to temperate notothenioids (Fig. 2a, Fig. S7) (Bista et al. 2023). The increase of young TEs in the LA region of cryonotothenioids and icefishes aligns with the genome-wide peak of young TEs observed in both red-blooded cryonotothenioids and icefishes compared to temperate notothenioid species, but the relative abundance of young TEs in the LA region was greater than genome-wide (Fig. S7 and Bista et al. (2023)). TEs and repeats (i.e. sequences “masked” by RepeatMasker [Smit et al. 2013]) account for 32 ± 5% of the whole genome assemblies in temperate notothenioids, while they account for 55 ± 5% of the genome assemblies in red-blooded cryonotothenioids and 60 ± 1% in icefishes (Fig. S8a). Masked sequences account for 30 ± 12% of the LA region in temperate notothenioids, similar to the proportion observed genome-wide (32 ± 5%, Fig. S8a and b). In contrast, masked sequences account for an astounding 78 ± 3% of the LA region in red-blooded cryonotothenioids and 84 ± 3% in icefishes (Fig. S8b). Genome wide, the increase in TEs and repeats in red-blooded cryonotothenioids and icefishes compared to temperate notothenioids appears overall proportional for each TE class, with only a notable increase in the LTR fraction (7 ± 2%, 12 ± 2%, and 16 ± 1% in temperate notothenioids, red-blooded cryonotothenioids, and icefishes, respectively) and a reduction in the unclassified class fraction (23 ± 5%, 18 ± 2%, and 16 ± 0% in temperate notothenioids, red-blooded cryonotothenioids, and icefishes, respectively) (Fig. S8d, Table S3). In the cryonotothenioid and icefish LA regions, the increase in TEs and repeats was mainly characterized by the proliferation of SINE, DNA, LINE, and unclassified elements (Fig. S8e).

Fig. 2.

LA region composition. a) Repeat landscape plots showing the distribution of TEs as a percentage of divergence from the consensus repeat models (Kimura divergence) versus their coverage of the LA region. Colors represent different TE classes. b) mVISTA alignment of the hbala2 gene using the Antarctic dragonfish G. acuticeps as reference and all icefish species. c) mVISTA alignment of 3,000 bp from the flanking gene aqp8.2 (not shown, on the right side) and containing the remaining hbala2 exon3 using the Blackfin Icefish C. aceratus as reference. In b) and c) hbala2 exons are colored in blue, conserved noncoding sequences in red, tRNAs in yellow, and TEs in green, pink, and cyan. Vertical dashed lines mark the ends of sequences to differentiate from an absence of conservation. d) Schematic representation of the overall LA cluster organization and surrounding TEs and tRNAs in red-blooded notothenioids and icefishes.

To test whether TEs were associated with the multiplication of tRNAs in the LA region, we merged the tRNA and TE annotations for the LA region. Results revealed that tRNAs in cryonotothenioids are often organized in triplets of tRNA-Ile, tRNA-Thr, and tRNA-Thr, usually preceded by an unclassified repeat element that RepeatMasker labeled “rnd-6_family-24817#Unknown” (hereafter “fam-24817#Unk.”) (supplementary material S2 and S3). This tRNA triplet and associated TE appears repeated almost perfectly numerous times in all cryonotothenioid species, including in icefishes, but was absent in temperate notothenioids (supplementary material S2 and S3).

When specifically looking at TEs present in the vicinity of the remaining α-la2 exon in icefishes, a pattern of associated tRNAs and TEs was strikingly conserved across species with only minimal variations. This pattern consisted of five elements: a fam-24817#Unk. element located upstream of a tRNA-Ile, followed by an rnd-1_family-1061#LINE/L2 element (hereafter fam-1061#LINE/L2) adjacent to an unclassified DR0081120#Unknown element (hereafter DR0081120#Unk.), and terminated by one or multiple tRNA-Thr (supplementary material S2 and S3). This “tRNA-TE unit” (fam-24817#Unk.–tRNA-Ile–fam-1061#LINE/L2–DR0081120#Unk.–tRNA-Thr) was present in a single copy in C. gunnari and P. georgianus, while it appeared twice with minimal variations in C. esox and four times in C. aceratus (supplementary material S2 and S3). No “tRNA-TE unit” could be found in the close vicinity of α-la2 in red-blooded cryonotothenioids and temperate notothenioids (supplementary material S2 and S3). We thus conclude that the “tRNA-TE unit” inserted near the α-la2 gene in an ancestor of all icefishes after its lineage diverged from dragonfishes.

To further query sequence conservation in the vicinity of the α-la2 gene across notothenioid species, we generated a ∼3 kb-long VISTA alignment (Frazer et al. 2004) with five red-blooded notothenioids and all 16 icefish species using genome assemblies alongside previously published icefish hemoglobin sequences (Near et al. 2006). Plots using the dragonfish G. acuticeps LA cluster sequence as reference showed that the DNA sequence is well conserved across all species from the end of the flanking aqp8.2 gene (right border of the alignment in Fig. 2b) until sequence conservation abruptly stops at the same position in all icefish species just before the α-la2 exon-3 (Fig. 2b). Aligning the same sequences using the icefish C. aceratus as reference instead of G. acuticeps revealed that the beginning of the fam-24817#Unk. element starting the “tRNA-TE unit” corresponds to the end of the alignment with red-blooded species and was precisely located at the same distance (i.e. 136 bp) from α-la2 exon-3 in almost all icefish species (Fig. 2c). Across icefish species, the alignment remains well conserved for most of the “tRNA-TE unit” where sequence is available (Fig. 2c). Notably, the sequence alignment of Jonah's Icefish Neopagetopsis ionah, which possesses a more complete, although nonfunctional hemoglobin cluster (Near et al. 2006), was more similar to red-blooded cryonotothenioids than to icefishes (Fig. 2b and c). This puzzling conservation could have occurred either by hybridization of the N. ionah ancestor with a red-blooded species (Near et al. 2006) or by incomplete lineage sorting, but resolving that ambiguity would require further analyses beyond the scope of this study. Nonetheless, these VISTA alignment analyses enabled the conclusion that the translocation of this “tRNA-TE unit,” only 136 bp away from α-la2 exon-3, occurred in the icefish ancestor after its lineage diverged from dragonfishes, and hence is likely associated with the disruption of hemoglobin genes.

We then examined the origin of this “tRNA-TE unit” and screened the LA region of red-blooded notothenioids for such a pattern. While we did not find this unit in the LA region of the temperate species C. trigloides and E. maclovinus (supplementary material S2 and S3), we successfully identified one or more duplicates of the tRNA-TE unit in all red-blooded cryonotothenioids about 20 to 30 kb away from the start of the α-la2 gene (supplementary material S2 and S3), in contrast to the 136 bp found in icefishes. Because both the chrn-like pseudogenes and the “tRNA-TE unit” are absent from temperate notothenioid LA regions but present in cryonotothenioid LA regions, we searched for the presence of the “tRNA-TE unit” in the vicinity of the additional chrn-like cluster located on the same chromosome but ∼11 Mb away from the LA region; search results failed to identify the unit or any of its components (supplementary material S3). We thus conclude that the “tRNA-TE unit” and its components translocated to the LA region independently of the translocation of the chrn-like pseudogenes.

A VISTA alignment using ∼40 kb-long sequences from red-blooded notothenioids and the dragonfish G. acuticeps as reference confirmed that sequence conservation in all icefish species re-aligns with red-blooded species on the first tRNA-TE unit located ∼25 kb away from α-la2 exon-3 (Fig. S9a). Reciprocally, the same VISTA alignment using instead the icefish C. aceratus as reference revealed a striking conservation between icefishes and red-blooded cryonotothenioids, with the exception that ∼27 kb of red-blooded sequence was absent in the alignment (Fig. S9b). Together, this conserved and precisely delimited “gap” in the icefish DNA sequence revealed that the region between the start of the fam-24817#Unk TE next to exon-3 of α-la2 in icefishes and the “tRNA-TE unit” ∼25 to 30 kb away in red-blooded notothenioids has seemingly been deleted from the icefish ancestor genome (Fig. 2d). We do not know the precise molecular mechanism leading to this “tRNA-TE unit” duplication and translocation from these specific TEs; however, our analysis precisely identified, to the base pair, the exact region containing the LA hemoglobin genes that was deleted in an ancestor of icefishes. Our results thus incriminate the mobility of TEs and the multiplication of tRNAs in the deletion of hemoglobin genes of the LA cluster in an icefish ancestor after its divergence from dragonfishes.

A Model for the Evolution of the LA Cluster and LA Region in Notothenioids

We therefore propose here a model retracing the steps of the LA cluster and LA region evolution in notothenioids (Fig. 3). In Perciformes, including the notothenioid ancestor (Fig. 3a), the LA cluster was composed of two alpha- and one beta-globin genes and multiple tRNAs located between rhbdf1b and the LA cluster, as well as, in some cases, a few tRNAs interspersed between globin genes or their exons. In the Eleginopsioidea ancestor (Fig. 3b), the β-la1 gene became tandemly duplicated, followed by the independent translocation of the chrn-like pseudogene cluster and of the “tRNA-TE unit” in the cryonotothenioid ancestor (Fig. 3c). Among red-blooded cryonotothenioid lineages (Fig. 3d), the dramatic amplification of tRNA genes already present in the vicinity of the LA cluster in Perciformes induced major changes in the size and organization of the region and may have caused the loss of some chrn-like pseudogenes through the reshuffling of the region in a lineage-specific manner. Similarly, the β-la1.2 gene duplicate was independently and repeatedly lost in various red-blooded cryonotothenioid lineages, and the tRNA-TE unit independently duplicated. Finally, in an icefish ancestor, after its lineage diverged from dragonfishes (Fig. 3e), the LA cluster was deleted concomitantly with the transposition of the tRNA-TE unit only 136 bp away from the exon 3 of α-la2. In summary, this genomic analysis revealed the extreme variability of the LA region in cryonotothenioids and is compatible with all predictions from the hypothesis of a TE-driven mechanism by which the LA cluster was deleted from the icefish ancestor genome after its divergence from dragonfishes.

Fig. 3.

Model of LA cluster and region evolution in notothenioids. Refer to the main text for detailed explanations of the evolutionary steps a)–e). Alpha-globin genes (hbala) and beta-globin genes (hbbla) are represented in red and blue, respectively, along with genes flanking the clusters in grey, tRNA genes in yellow, chrn-like pseudogenes in pink, and the “tRNA-TE” unit in green. Pseudogenes are represented with half-filled squares. Genes above the line are transcribed toward the right, and genes below the line are transcribed toward the left. Divergence time estimates are from Bista et al. (2023).

Extreme Birth-and-Death Gene Evolution Shaped the MN Cluster in Red-Blooded Notothenioids

The MN cluster gene content is highly variable among notothenioid species (Fig. 4a, Table S4) (Bista et al. 2023; Desvignes et al. 2023), and more broadly among teleosts (e.g. Opazo et al. 2013; Hotaling et al. 2023; Omelchenko et al. 2024). In notothenioids, however, copy-number variations seemingly involve only the hbamn1 (α-mn1) and hbbmn1 (β-mn1) gene pairs, whereas in other Perciformes groups, both the α-mn1/β-mn1 gene pairs and hbamn2 (α-mn2) and hbbmn2 (β-mn2) gene pairs vary in copy number (Hotaling et al. 2023). As previously shown (Bista et al. 2023; Desvignes et al. 2023), some plunderfish and dragonfish species have independently lost α-mn2 or β-mn2 (Fig. 4a); notably, these two genes of the MN cluster have dramatically reduced expression in adult cryonotothenioids (Desvignes et al. 2023). In contrast to the LA cluster, no traces of alpha- or beta-globin genes can be found in the “MN region” of icefishes between the flanking genes kank2 and nprl3 (Fig. 4a) (Bargelloni et al. 2019; Kim et al. 2019; Bista et al. 2023; Desvignes et al. 2023).

Fig. 4.

MN cluster region organization. a) Phylogenetic relationships and schematic representation of the genomic organization of the MN cluster and neighboring genes. Alpha-globin genes (hbamn) and beta-globin genes (hbbmn) are represented along with genes flanking the clusters. Pseudogenes are represented with half-filled squares. Arrows for the entire cluster represent orientations in each assembly (forward toward the right, reverse toward the left). Genes above the line are transcribed in the same orientation as the cluster, while genes below the line are transcribed in the opposite direction to the cluster. Nucleotide phylogenetic trees of b) hbamn1 coding sequences and c) hbbmn1 coding sequences. Different species are color-coded to facilitate observing the grouping of sequences by species.

Phylogenetic trees of α-mn1 and β-mn1 genes revealed that paralogs within species group together (Fig. 4b and c), thus revealing that paralogs within species are more closely related to one another than they are to potential orthologs across species. This pattern can be explained by either of two nonmutually exclusive mechanisms: gene turnover (“birth-and-death evolution”), in which only one pair of α-mn1 and β-mn1 was retained in each lineage, diverged, and tandemly duplicated independently in each species by unequal crossing over; and/or by gene conversion, a form of nonreciprocal recombinational exchange between closely linked, usually tandemly duplicated genes in which a paralog's sequence is repaired according to a neighboring paralog's sequence after a double-strand DNA break, usually in meiosis, leading to the gradual homogenization of paralogous genes in a family, a pattern referred to as concerted evolution (Nei and Rooney 2005; Storz 2018; Lorenz and Mpaulo 2022). Both of these mechanisms have shaped the evolution of hemoglobin genes in various vertebrate species (e.g. Runck et al. 2009; Storz et al. 2011, 2012; Hardison 2012; Storz 2018; Signore et al. 2019), including in human with numerous cases of gene conversion and cases of unequal crossover leading to multiplications or deletion of α-globin and γ-globin genes, and pathological conditions such as hemoglobin Lepore, hemoglobin anti-Lepore, and hemoglobin Kenya (e.g. Borg et al. 2009; Neumann et al. 2010; Ruhl et al. 2022). In notothenioids, metallothioneins have also been shown to have evolved by gene conversion (Bargelloni et al. 1999).

To test whether concerted evolution by gene conversion can explain the sequence similarities between α-mn1 and β-mn1 gene paralogs in notothenioids, we performed independent phylogenetic analyses for different portions of the genes (i.e. each of the three exons, each of the two introns, 300 bp upstream of the start codon, and 300 bp downstream of the stop codon [supplementary material S4]). Because concerted evolution more frequently affects protein-coding parts of the genes (i.e. exons) (Storz et al. 2007; Runck et al. 2009; Natarajan et al. 2015; Storz 2018), phylogenetic tree patterns for coding and noncoding partitions should be inconsistent. Except for a few rare exceptions attributable to long-branch attractions and unusually rapid divergence of one paralog partition, results revealed that for both α-mn1 and β-mn1 genes, each partition of each gene displays similar topologies with paralogs grouping together (Fig. S10). We thus conclude that gene conversion did not shape the evolution of α-mn1 and β-mn1 genes in notothenioids; hence, that birth-and-death evolution, characterized by gene losses and tandem duplications in each lineage, a process similar to pathological situations known in humans (e.g. Borg et al. 2009; Neumann et al. 2010; Ruhl et al. 2022), is the most likely mechanism explaining the sequence similarity of α-mn1 and β-mn1 paralogous genes.

Because copy number variations by tandem duplications in paralogous genes are mainly caused by unequal crossing-over during recombination between partially misaligned chromosomes in meiosis, we examined the presence of repeated sequences and TEs surrounding α-mn1 and β-mn1 gene pairs. Repeats and TEs can facilitate the misalignment of chromosome segments during meiosis and thus the occurrence of unequal crossing over, leading to duplications and deletions in hemoglobin genes. In contrast to the LA region, the MN region did not experience a dramatic increase in the abundance of TEs in red-blooded cryonotothenioids compared to temperate species, and TE abundances are overall low in the MN region (Fig. 5a, Figs. S7 and S8). Self-alignments and species-pair alignments of the MN regions in red-blooded species also failed to reveal patterns of major repeated sequences, except for the duplicated α- and β-globin genes (Fig. 5b1 to 2, Figs. S11 to S13).

Fig. 5.

MN region composition. a) Repeat landscape plots showing the distribution of TEs as a percentage of divergence from the consensus repeat models (Kimura divergence) versus their coverage of the MN region. Colors represent different TE classes. b) Dot plots of self-alignments for the MN region and flanking genes and c) dot plots of species pair alignments. Green and red represent forward and reverse alignments, respectively. Grey, blue, orange, green, and red shadings highlight the flanking genes, hbamn1/hbbmn1 genes, hbamn2/hbbmn2 genes, DNA Kolobok-T2 TEs, and the MN-TE unit, respectively. d) Schematic representation of the overall MN region organization in red-blooded notothenioids and icefishes.

Combining the TE and hemoglobin gene annotations in the MN regions for each species, however, revealed that α-mn1 and β-mn1 gene pairs are generally surrounded by similar TEs within species but that those TEs are rarely conserved across species (Table S4, supplementary material S3). For example, in the temperate notothenioid C. trigloides, the α-mn1 and β-mn1 gene pairs are generally surrounded by DNA/hAT-Ac TEs (DR0220495, DR0218602, and DR0217532), and unclassified TEs (DR0271023) and DNA/Zisupton TEs (DR0112103). In the Antarctic toothfish D. mawsoni, the α-mn1 and β-mn1 gene pairs are surrounded by DNA/hAT-Ac TEs (DR0150463 and DR0217532) and unclassified TEs (DR0258035). And in T. bernacchii, the α-mn1 and β-mn1 gene pairs are surrounded by DNA/hAT-Ac TEs (DR0063726, DR0219577, and DR0271425), unclassified TEs (DR0258035), LINE/L2 TEs (DR0257416 and DR0151171), and RC/Helitron TEs (DR0211913). In some species, certain TEs are nonetheless shared in the surroundings of α-mn1 and β-mn1 gene pairs (e.g. DNA/hAT-Ac TEs DR0217532 in C. trigloides and D. mawsoni; unclassified TEs DR0258035 in D. mawsoni and T. bernacchii); however, most frequently, TEs are not shared between species, and inspecting their sequences did not provide evidence that they would be evolutionarily related (supplementary material S3). This nonconservation of TEs in the direct vicinity of α-mn1 and β-mn1 gene pairs across species is consistent with the hypothesis that different TEs facilitated unequal crossing-over independently in each lineage and led to species-specific hemoglobin gene pair amplification and contractions.

TE-Mediated Deletion of the MN Hemoglobin Gene Cluster in the Icefish Ancestor

In contrast to the MN regions in red-blooded notothenioids, the MN regions in icefishes show a drastic increase in young TE and repeat abundance, primarily characterized by LINE, DNA, and unclassified elements (Fig. 5a, Fig. S7). The self-alignments of icefish MN regions, however, do not show a pattern of sequence repetitions (Fig. 5B3, Fig. S11). This evidence suggests that the MN region is composed of unrelated sequences instead of repeated sequences such as the many tRNAs observed in the LA region (Fig. 1b). Furthermore, dot-plot alignments of icefish MN regions to red-blooded notothenioids failed to reveal sequence conservation within the expected position of the MN cluster and its hemoglobin genes in red-blooded species (Fig. 5c1 to 2, Figs. S12 and S13). Alignments between icefish species pairs also show a lack of conservation in the central part of the MN region (Fig. 5c3, Fig. S14). Thus, the portion of the MN region in icefishes where hemoglobin genes used to be is highly variable across icefish species. This finding shows that this portion of the MN region continued to evolve in a lineage-specific manner after hemoglobin gene deletions in icefishes.

Detailed inspection of the ends of sequence alignments between red-blooded and icefish species, however, provided clues on the process by which hemoglobin genes may have been lost from the genome of the icefish ancestor. The kank2-side of the MN region has a pair of DNA/Kolobok-T2 TEs (DR0131494#DNA/K-T2 and DR0066537#DNA/K-T2) shared by all studied notothenioid species, but alignments between red-blooded species and icefishes terminated abruptly and consistently exactly 241 bp after these TEs (Fig. 5c1 to 2, Figs. S12 to S14, Table S4, supplementary material S3). In contrast, alignments between icefish species consistently continue for approximately 6 kb after the DNA/Kolobok-T2 TE pair (Fig. 5c3, Fig. S14). Annotation of TEs and repeats in this icefish-specific extended alignment revealed that it is composed almost entirely of TEs. Importantly, this TE-rich portion of the MN region appears conserved in all icefishes studied, but it is absent in red-blooded species. This region, hereafter referred to as the “MN-TE unit,” starts precisely where the alignment with red-blooded species ends, which corresponds to the end of a DNA/hAT-Tip100 (DR0081006#hAT-Tip100) element and ends precisely where the alignment between the various icefish species ends, which corresponds to the end of a LINE/L2 element (DR0081282#LINE/L2) (Fig. 5d). This MN-TE unit contains multiple densely organized and often overlapping TEs, including the previously mentioned DNA/hAT-Tip100 (DR0081006#DNA/hAT-Tip100), a LINE/RTE-X (DR0082231#LINE/RTE-X), two different unclassified elements (rnd-1_family-1014#Unknown and rnd-1_family-1002#Unknown), two DNA/P elements (DR0081438#DNA/P and rnd-1_family-166#DNA/P), a DNA/IS3EU element (DR0152203#DNA/IS3EU), and ends by the previously mentioned LINE/L2 element (DR0081282#LINE/L2) (Fig. 5d). On the nprl3-side, no TE or repeat shared by all notothenioids could be identified. However, the start of a DNA element (rnd-1_family-896#DNA, hereafter fam-896#DNA) located about 720 bp upstream of nprl3 and conserved in all studied icefish species corresponds exactly to the end of the alignment between red-blooded species and icefishes (Fig. 5c2 to 3 and d). A search for the TEs composing the MN-TE unit and the fam-896#DNA TE failed to identify these elements anywhere in the MN region of any studied red-blooded species. Hence, we conclude that the entire portion of the MN region that contained hemoglobin genes in the red-blooded icefish ancestor after its divergence from dragonfishes was deleted all at once, simultaneously with the insertion of a genomic fragment ending on one side by the MN-TE unit and on the other side by a fam-896#DNA element. Following icefish species divergence, the region between the MN-TE unit and the fam-896#DNA element further diversified in different icefish lineages, leading to the loss of sequence conservation across species. It is thus unlikely that the deletion of the MN cluster in the icefish ancestor occurred by a local unequal crossing-over event driven by the presence of repeated sequences, the way we hypothesized α-mn1 and β-mn1 gene pairs duplicated in each red-blooded species. Instead, the deletion may have occurred by another mechanism, such as during TE mobilization and insertion into the MN region or an ectopic recombination event with another part of the chromosome or genome (Bourque et al. 2018; Wells and Feschotte 2020; Balachandran et al. 2022).

A Model for the Evolution of the MN Cluster and MN Region in Notothenioids

Together, these observations suggest the model shown in Fig. 6, which retraces evolutionary steps for the MN cluster and MN region in notothenioids. In the notothenioid ancestor (Fig. 6a), the MN cluster was composed of a single pair of α-mn2 and β-mn2 genes and likely multiple pairs of α-mn1 and β-mn1 genes downstream of a pair of Kolobok-T2 DNA TEs. In each lineage and each species (Fig. 6b), birth-and-death evolution characterized by extreme gene turnover of α-mn1 and β-mn1 gene pairs led notothenioids to display various numbers of α-mn1 and β-mn1 gene pairs, some potentially becoming pseudo-genes. In plunderfishes and dragonfishes (Fig. 6c), α-mn2 and β-mn2 genes may have been lost or pseudogenized independently in different lineages and species. Finally, in the icefish ancestor after its divergence from dragonfishes about 5.3 MYA (Fig. 6d), the MN cluster was deleted concomitantly with the likely simultaneous insertion, between the Kolobok-T2 TE pair and the flanking gene nprl3, of a genomic fragment precisely ending on one side by the MN-TE unit and on the other side by a fam-896#DNA element.

Fig. 6.

Model of MN cluster and region evolution in notothenioids. Refer to the main text for detailed explanations of the evolutionary steps a)–d). Alpha-globin genes (hbamn) and beta-globin genes (hbbmn) are represented in red and blue, respectively, along with genes flanking the clusters in grey, DNA Kolobok-T2 TEs in green, rnd-1 TE in pink, and the “MN-TE” unit in Salmon. Pseudogenes are represented with half-filled squares. Genes above the line are transcribed toward the right, and genes below the line are transcribed toward the left. “Lineage 1” and “Lineage 2” provide examples of how differential birth-and-death evolution may have differed between taxa.

In summary, analysis of the structure of genomic repeats revealed that the extreme variability of the MN gene cluster in notothenioids was likely facilitated by lineage-specific TEs associated with alpha- and beta-globin gene pairs and a proposed TE-driven mechanism by which the MN cluster was deleted from the icefish ancestor genome after its divergence from dragonfishes about 5.3 MYA (Bista et al. 2023).

Conclusions

Antarctic white-blooded icefishes are evolutionary oddities that have fascinated biologists for decades because they lack hemoglobin, a unique phenotype among vertebrates. Previous work identified the remnant of a single hemoglobin gene in all but one icefish species (Near et al. 2006) and revealed that hemoglobin genes were under diversifying selection during cold acclimation and have since been pervasively diversifying compared to their temperate relatives (Desvignes et al. 2023). Gene expression analyses across notothenioids further uncovered the genetic bases of reduced hemoglobin multiplicity in Antarctic cryonotothenioids and suggested that the icefish ancestor likely also relied on a single pair of alpha- and beta-globin genes, making it vulnerable to any detrimental mutation affecting hemoglobin function and potentially predisposing their loss (Desvignes et al. 2023). Analyses of hemoglobin gene sequences, however, failed to reveal mechanisms by which or positive selective pressures favoring hemoglobin gene removal from the icefish ancestor genome.

In the present study, using recently generated high-quality notothenioid genome assemblies (Kim et al. 2019; Bista et al. 2020, 2023, 2024; Lee et al. 2021; Cheng et al. 2023; Jo et al. 2023; Rivera-Colón et al. 2023; Ryder et al. 2024), we were for the first time able to identify evidence consistent with specific genetic mechanisms by which hemoglobin gene clusters likely evolved across notothenioid species. Detailed comparison of hemoglobin genomic regions in red-blooded notothenioids and icefishes further suggested mechanistic models explaining how each of the two hemoglobin clusters was independently deleted from the genome of the icefish ancestor. TEs have been implicated in the evolution of the cryonotothenioid genome in size (Bista et al. 2023) and architecture (Auvinet et al. 2018, 2019, 2020), and present data incriminate TEs in facilitating the deletion of both hemoglobin clusters. For both clusters, TEs appear to have played an outsized role with the dramatic multiplication of tRNAs in the LA cluster vicinity and in the extreme birth-and-death evolution of the globin genes in the MN cluster (Figs. 3 and 6). In the icefish ancestor, the translocation of an ancient tRNA-TE genomic unit shared by all cryonotothenioids into the second intron of α-la2 can easily explain the simultaneous deletion of the LA cluster (Fig. 3). In contrast, the deletion of the MN cluster appears to have been driven by its replacement by a novel genomic fragment flanked by TE-rich sequences and extending from just a few hundred base-pairs upstream and downstream of the hemoglobin cluster (Fig. 6).

While these new results identify likely mechanisms by which hemoglobin clusters were independently deleted from the icefish ancestor's genome, the reasons why this unique and seemingly detrimental phenotype persisted in living icefishes instead of being selected against remain a mystery. Was it beneficial to lose hemoglobin genes? Was it because the overall depletion of iron in the Southern Ocean made the biosynthesis of hemoglobin too energetically expensive (Corliss et al. 2019; Laptikhovsky 2019)? Or was it because getting rid of hemoglobin and aborting the erythropoiesis process would make their blood thinner and then counteract the increased viscosity caused by the frigid temperatures of the Southern Ocean (Wells et al. 1990; Egginton 1996)? Or was hemoglobin simply dispensable given the increased oxygen content of cold water? Previous work showed that the Bullhead Notothen Notothenia coriiceps survives when its hematocrit is reduced by 90% and hemoglobin concentration is lowered by 70% following treatment with phenylhydrazine (Borley et al. 2010). The Emerald Notothen T. bernacchii can also survive several days when 95% of its hemoglobin is bound to toxic carbon monoxide (di Prisco et al. 1992). And the red-blooded Deep-water Dragonfish Bathydraco marri carries up to one-sixth of its oxygen dissolved in the blood plasma rather than bound to its hemoglobin in red blood cells (Kunzmann et al. 1991). Thus, if the icefish ancestor was already predisposed to cope with reduced hemoglobin function using alternative ways to transport oxygen, the loss of hemoglobin genes may not have been a lethal event and not strongly selected against due to the limited competition and predation risk in the Southern Ocean. While additional physiological and phenotypic studies of more Antarctic fish species are needed to better understand their capacity to live with little to no hemoglobin, results presented here strongly support a model for how hemoglobin genes were lost in the icefish ancestor and thus bring us closer to understanding the deep evolution of this iconic clade.

Materials and Methods

Genome Assemblies and the Annotation of Hemoglobin and tRNA Genes

Published genome assemblies of 40 notothenioids (Ahn et al. 2017; Baalsrud et al. 2018; Bargelloni et al. 2019; Kim et al. 2019; Bista et al. 2020, 2023, 2024; Catchen et al. 2020; Lee et al. 2021; Cheng et al. 2023; Jo et al. 2023; Rivera-Colón et al. 2023; Corso et al. 2024; Ryder et al. 2024) and 2 non-notothenioid Perciformes (Feron et al. 2020; Hotaling et al. 2023) were searched for hemoglobin genes as previously described in Desvignes et al. (2023). Briefly, LA and MN clusters were first located in each assembly by searching for their flanking genes (i.e. rhbdf1b and aqp8.2 for the LA cluster, and kank2 and nprl3 for the MN cluster) and the sequence, length, position, and strand of individual exons of each alpha- and beta-globin gene were manually retrieved from each assembly by performing BLASTN searches using the yellow perch P. flavescens as reference for exon boundaries. Exons of each gene were concatenated and translated into protein sequence to verify their accuracy. If a gene was incomplete (e.g. a verified missing exon) or displayed a premature stop codon, the gene was considered pseudogenized. Only genome assemblies providing highly contiguous sequences for the LA or MN clusters were retained for further analyses (Tables S1). Positions and orientation of all hemoglobin and flanking genes for the LA and MN clusters in the retained species are provided in Tables S2 and S4, respectively. Genes encoding tRNAs were annotated genome-wide and in the LA region with tRNA-scan SE 2.0 using default parameters (Chan et al. 2021); supplementary materials S1 and S2 provide tRNA gene annotations.

Dot Plots and Genomic Alignments

Genomic regions between the flanking genes on each side of hemoglobin clusters were retrieved from high-quality genome assemblies based on BLASTP results for the position of each cluster’s flanking genes. Sequence self-alignments and species-pair alignments (i.e. dot plots) were generated with YASS v1.16 using default parameters (Noé and Kucherov 2005), and multiple alignments were generated with mVISTA using the LAGAN alignment program with default parameters (Frazer et al. 2004). LAGAN alignments were annotated with each protein-coding gene exons, tRNA genes, and TEs in both the Antarctic dragonfish G. acuticeps and the Blackfin Icefish C. aceratus.

Phylogenetically Corrected Regressions

Because evolutionary relatedness can bias regression or correlation coefficients, regression analyses were performed using the Phylogenetic Generalized Least Squares (pGLS) model implemented in the R package caper v. 1.0.3 using Pagel's λ estimates from an ML least squares regression residuals. For each model, the likelihood surface of λ was visually inspected to avoid cases of local optima. Data and regressions were plotted using the function phylomorphospace in the phytools v.2.4-4 package.

TE Annotation

To annotate and mask repeat elements in the genome and in the two hemoglobin clusters, we used a curated library of genomic repeats first described by Rivera-Colón et al. (2023). Briefly, this repeat library combines the annotations present in the teleost-specific partition of Repbase release 27.02 (Bao et al. 2015) with de novo repeat annotations of several notothenioid genomes, as generated by RepeatModeler version 2.0.2a (Flynn et al. 2020). For each genome, we annotated and masked repeat sequences using RepeatMasker version 4.1.5 (Smit et al. 2013). We specified our custom repeat library using the -lib option, exported the alignments of annotated elements (-a), returned soft-masked sequences (-xsmall), and an annotation of masked regions in GFF format (-gff). Following annotation and masking of repeats, we calculated Kimura divergence of all elements using the calcDivergenceFromAlign.pl utility script from RepeatMasker version 4.1.5, specifying the repeat alignments as input. This process was repeated separately for the sequences of both MN and LA clusters for all species. Additional parsing of the RepeatMasker outputs was performed using custom Python scripts (supplementary material S3).

Phylogenetic Trees

For each hemoglobin gene in the MN cluster, independent phylogenetic trees were produced for each of the three exons, the two introns, the 300 bp upstream of the first exon, the 300 bp downstream of the third exon, and for the full coding sequence (i.e. concatenated exons) using IQ-TREE v.1.6.12 (Trifinopoulos et al. 2016; Minh et al. 2020). When needed, nucleotide sequences were aligned using MAFFT 7.520 (Katoh et al. 2019) and the best-fit model was determined using ModelFinder based on the Bayesian information criterion (Kalyaanamoorthy et al. 2017). Consensus trees were constructed from 1,000 Ultrafast bootstrap trees (Hoang et al. 2018), and branch lengths were tested using a SH-aLRT branch test using 1,000 replicates (Guindon et al. 2010). All alignments, ModelFinder, and IQ-TREE results are available in supplementary material S4.

Supplementary Material

Supplementary material is available at Genome Biology and Evolution online.

Author Contributions

Conceptualization: T.D. and J.H.P.; Data curation: T.D. and A.R.C.; Formal analysis: T.D. and A.R.C.; Funding acquisition: J.H.P. and T.D.; Investigation: T.D. and A.R.C.; Methodology: T.D. and A.R.C.; Project administration: T.D. and J.H.P.; Resources: T.D., A.R.C., and J.H.P.; Software: T.D. and A.R.C.; Supervision: T.D.; Validation: T.D. and A.R.C.; Visualization: T.D.; Writing—Original draft preparation: T.D.; and Writing—Review & editing: T.D., A.R.C., and J.H.P.

Funding

This work was supported by the National Science Foundation Office of Polar Program (OPP-1947040, OPP-1543383, and OPP-2232891 to J.H.P. and T.D.). This work benefited from access to the University of Oregon high-performance computer Talapas (NSF grant OCI-0960354).

Data Availability

All data generated or analyzed during this study are included in the published article (and its Additional Information files), are publicly available in NCBI, or are deposited in the dedicated United States Antarctic Program Data Center (USAP-DC) project repository p0010417 (https://www.usap-dc.org/view/project/p0010417).