A new phased assembly of the Antarctic spiny plunderfish provides novel insights into the evolution of the notothenioid radiation

13

14 Notothenioids are a well characterised species flock endemic to the Antarctic and an important model 15 group for the study of genome adaptation to extreme cold. We used a new reference assembly and clade-16 wide comparative genomic analysis to investigate cryonotothenioid evolution and the appearance of 17 novel functionalities linked to cold adaptation. 18

19 A new phased assembly of a model notothenioid, Harpagifer antarcticus, demonstrated low levels of 20 haplotypic variability across the genome . Nevertheless, numerous insertions from multiple LINE -L2 21 clades were found, suggesting ongoing transposition with potential contribution to speciation. Contrary 22 to expectations the afgp locus was highly similar between haplotypes, except for large length allelic 23 variants of afgp genes. Analysis suggests a model for the afgp locus expansion in H. antarcticus through 24 segmental tandem duplications involving two pairs of afgp genes at time. Syntenic reconstruction of 25 genomes from across the clade demonstrates conserved macrosyntenic relationships and group specific 26 chromosomal fusions of notothenioids. Quantification of genome gain and transposition rates during 27 cryonotothenioid diversification showed a first ancestral slow genome expansion concurrent with 28 historic temperature drops. This was followed by lineage-specific massive peaks of genomic gain and 29 transposition activity . Finally, w e identified a set of genes that underwent ancestral diversifying 30 selection and acquired novel conserved non -coding elements during the cryonotothenioid emergence. 31 These were related to antioxidants and proteostasis , which may have facilitated the notothenioid 32 Antarctic radiation. 33

34 Diversifying selection and genomic gain linked to transposon activity are primary contributors to 35 lineage-specific evolutionary dynamics through the clade which facilitated adaptation to life in the cold. 36

37 cold adaptation, transposons, LINE-L2, afgp, haplotype diversity, genome expansion, proteostasis, 38 diversifying selection, antioxidant activity 39 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 2

40 The notothenioids are one of the most diverse and dominant fish groups in the Southern Ocean and a 41 rare example of a well characterised marine flock species [1]. The clade is formed by approximately 42 140 species grouped in eight families, including three non-Antarctic (basal groups) and five mostly high 43 specialised Antarctic families which are also known as cryonotothenioids [2]. They are characterised 44 by genomic adaptations to extreme cold and they are considered to be an important model group for the 45 study of cold adaptation [3]. 46 The appearance of the antifreeze glycoproteins (AFGPs) is one of the most characteristic innovations 47 enabling freeze avoidance in this group, and are present in members of the cryonotothenioid clade [4]. 48 However, multiple other adaptations on the molecular and cellular level have also been considered to 49 play important roles in the survival and establishment of notothenioids (e.g., cold stable enzyme 50 production, lipid composition of cell ular membranes, modified mechanisms of protein homeostasis) 51 (summarised in [5]). The mechanisms through which such adaptations may have arisen, or the timing 52 of their appearance along the ancestral cryonotothenioid clade remain poorly understood. 53 Assembly of notothenioid genomes has been confounded by the high number of repeat elements present 54 in these genomes [6]. This is exemplified by t he structure of the antifreeze gene locus , which is 55 characterised by tandemly arranged genes [4,7], and high levels of repeat content making it very 56 challenging to assemble and analyse accurately [6]. For the afgp gene locus in particular it has been 57 suggested that high levels of structural variation on the haplotype level may be present. These estimates 58 were based on BAC library sequencing reconstruction of the afgp locus of the Antarctic toothfish 59 species Dissostichus mawsoni [8]. In that case it was shown that the two haplotypes of the afgp locus 60 were characterised by extensive structural variation, with a distinct number of afgp gene copies, with a 61 difference in overall size and mapping within each haplotype of the locus. Other reconstructions of the 62 afgp locus based on whole genome sequencing data were only able to map the consensus structure of 63 one assembled haplotype [6,9,10] though haplotig data have also been analysed [6]. To be able to test 64 this hypothesis on other species, haplotype resolved (phased) genome assemblies are required. 65 Using Oxford Nanopore Technologies (ONT) and Hi-C data we assembled a new haplotype resolved 66 chromosomal reference genome assembly for Harpagifer antarcticus (Antarctic spiny plunderfish). 67 This is the first phased assembly for any notothenioid species . In addition, H. antarcticus, as a small, 68 shallow-water and experimentally tractable member of the cryonotothenioids, is also an important 69 model species for the study of cold adaptation [11–13]. 70 Previous work on notothenioid genomes revealed the significant contributions of transposable elements 71 (TEs) in the evolution of the clade. For the cryonotothenioid clade in particular it has been shown that 72 a recent massive burst of transposons was linked to a significant genome expansion compared to the 73 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 3 sub-Antarctic group [6,14,15], while the evolution of various other genomic features has also been 74 linked to transposon activity such as expansion of the antifreeze locus, and the loss of haemoglobins in 75 the Channichthyidae family [6]. Furthermore, it has been suggested that TE accumulation in some 76 notothenioid groups, may have promoted karyotypic reorganisations likely leading to species 77 diversification [16]. 78 In general, TEs are known to be a major source of genomic variation across eukaryotes by disrupting 79 gene structures , reshaping regulatory networks, and co-opting novel regulatory elements [17]. The 80 extensive presence of TEs throughout the genome might amplify the occurrence of recombination 81 events thus promoting deletions, inversions, and translocations [18], while the presence of 82 environmental stressors may also promote TE activity [19]. Using whole genome comparisons can be 83 a powerful method for understanding the mechanisms and patterns of transposon activity at a large scale, 84 while it could also enable highly accurate quantification of genomic losses and gains occurring during 85 diversification events [20–22]. 86 In the present study we employed large scale genome wide analyses of chromosomal references across 87 the notothenioid radiation along with novel genomic resources, to understand the 1) overall contribution 88 of transposable elements in the cryonotothenioid diversification, through timed and quantified analysis 89 of genomic gains and losses, 2) the evolutionary dynamics of the antifreeze gene locus across the 90 cryonotothenioids, and models of expansion of afgps and haplotypic diversity in the H. antarcticus 91 phased assembly, 3) impact of other genomic features that may have enabled genomic adaptation to 92 extreme conditions. 93 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 4

94 Genome assembly and gene annotation 95 The assembled genome for H. antarcticus using ONT long reads and Hi-C data resulted in a fully phased 96 assembly with two high-quality haplotypes (Fig. 1, Fig. S1). The genome was assembled at a total length 97 of 1,069 Mb (Hap1) and 1,222 Mb (Hap2), and contig N50 of 17 Mb (Hap1) and 12,8 Mb (Hap 2). Each 98 haplotype was assigned to 24 chromosomes (Fig. 1A, C) with a scaffold N50 of 44.8 Mb (Hap1) and 41 99 Mb (Hap2) (Table S1). Both haplotypes were highly complete, with 99% BUSCO completeness for both 100 haplotypes, and QV scores of QV 40 (Hap 1) and QV 40.4 (Hap2) ( Table S1). Chromosomes were 101 named based on syntenic relationships with Cottoperca gobio genome assembly (fCotGob3.1) [23] (Fig. 102 1C; Table S2). Completeness of the assembly was confirmed by a clear enrichment of telomeric repeats 103 on at least one end of all chromosomes on both haplotypes (Fig. 1A, Fig. S2), confirming previous higher 104 level localisation data using in situ hybridization data of the (TTAGGG)n repeat motif [24]. Based on k-105 mer analysis, the heterozygosity rate of the H. antarcticus genome was estimated at 0.37%. 106 107 Fig. 1: Genome assembly and genomic features. A. Circos plot showing the 24 chromosomes of the 108 fHarAnt1.2 genome assembly (Hap1) and the distribution of protein coding genes (Genes), transposable 109 elements (TEs), tandem repeats (TR), and canonical telomeric repeats (TeloRep) . Chromosomes are 110 ordered by length and named based on syntenic positions with Cottoperca gobio reference genome [23]. 111 B. Photograph of a H. antarcticus specimen from Ryder Bay, Antarctic Peninsula (Photo credit: BAS 112 Photo Library. Photographer: Simon Brockington) . C. Hi-C-contact map of fHarAnt1. 2 Hap1. D) 113 Syntenic relationships between H. antarcticus Hap1 and C. gobio genomes. 114 De-novo gene annotation (BREAKER) for Hap1 identified 23,057 protein-coding genes, with an average 115 length of 14,904 bp, an average of 8.3 introns per gene, and mean intron and exon lengths of 1,555 bp 116 and 174 bp, respectively (Table S3). By lifting Hap1 annotation onto Hap2 we identified 22,463 (Table 117 S3) protein coding genes with 696 flagged as pseudogenes (i.e., containing premature stop codons). 118 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 5 BUSCO completeness of the gene annotation was high for both haplotypes with 97 .7% for Hap1 and 119 95.6% for Hap2 (Table S3). 120 Genomic landscape of repetitive elements 121 122 Fig. 2: LINE -L2 elements in the H. antarcticus genome. A. Repeat landscape profile showing the 123 activity of the three main TE groups through absolute time. Each bin represents one million years (MYA 124 = million years ago). B. Maximum likelihood phylogenetic tree of clustered LINE -L2 rever se-125 transcriptase sequences (fHarAnt1.2 genome). Tips represent the longest element for each cluster. The 126 size and colour of bubbles is proportional to the number of elements contained in each cluster. Arrows 127 highlight the eight identified high -copy number families. Tips without bubbles represent reference 128 sequences extracted from [25]. C. Genome coverage per LINE -L2 family, and D. activity through 129 absolute time of the eight LINE-L2 families highlighted in panel B. Each bin represents 500 thousand 130 years. E. Number of insertions across notothenioid genomes for each LINE-L2 family identified in H. 131 antarcticus. 132 Annotation of transposable elements on the fHarAnt 1.2 genome revealed 41% coverage (Total repeats 133 43%) (Table S4) with a clear enrichment towards chromosome ends (Fig. 1A). Each of the three major 134 TE classes (LINEs, LTRs, DNA), were almost equally represented in Hap1, covering approximately 13% 135 of the genome respectively , whereas LTRs are more prevalent on Hap2 (17.84% of the genome) . 136 Classification at the RepeatMasker superfamily level was achieved for just 32% of LTR and 41% of 137 DNA elements, in contrast to 97% of LINE elements. These predominantly consisted of LINE-L2 which 138 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 6 accounted for 9.6% (Hap1) and 8.44% (Hap2) of the genome ( Table S4). Tandem repeats and simple 139 repeats accounted for about 4.5% and 2% of the genome on the haplotypes, respectively. Repeat 140 landscape profiles describing activity of TEs through absolute time revealed that the TE content of the 141 Antarctic spiny plunderfish genome mainly derived from a recent burst of activity of all three major TE 142 groups that occurred after the diversification of cryonotothenioids, between 4 and 7 Million Years Ago 143 (MYA) (Fig. 2A). 144 To further characterize the evolutionary history of the highly abundant LINE -L2 elements we first 145 employed a phylogenetic approach to identify evolutionarily distinct LINE-L2 families ( Fig. 2B). 146 Among the 33,670 insertions harbouring a reverse transcriptase segment, 86% clustered with reference 147 L2 elements, whereases others are part of other L2-related clades (Crack, Daphne L2A and L2B) [25]. 148 Eight different high copy-number LINE-L2 families were identified, which were named HarAnt_LINE-149 L2_I – VIII. These families accounted for 60% of the total LINE -L2 content, and 57% of the genome 150 (both haplotypes). The two most abundant of these were L2_IV (27.4% LINE-L2, 1.5% of the genome), 151 and L2_II (22.9% LINE-L2,1.3% of the genome) (Fig. 2C). Dated repeat landscape plots suggest that all 152 families were already active before the extant cryonotothenioid radiation, with a first peak concurrent 153 with their emergence time ~10 MYA and a second more recent activity peak between 4 MYA and 500 154 thousand years ago (Fig. 2D). The two most abundant appear to have decreased their transposition rate 155 over the last 1 MYA, whereas the less abundant ones LINE-L2_VI, LINE-L2_VII, and LINE L2_VIII, 156 have become predominant. 157 Subsequent analysis of LINE -L2 families across 12 notothenioid genomes (Table S5), identified all 158 LINE-L2 families across all cryonotothenioid species. The three most abundant LINE-L2 families in the 159 H. antarcticus genome were also the most abundant ones in the other species ( Fig. 2E). In the sub -160 Antarctic species E. maclovinus and C. gobio, some low-abundance families, 2 and 3 respectively, were 161 missing and therefore likely have originated along the cryonotothenioid stem branch (Fig. 2E). 162 A total of 170 (18S), 534 (5S), 166 (5.8S), and 181 (28S) ribosomal RNA (rDNA) genes were annotated 163 in the H. antarcticus genome. Among the 18S, 5.8S, and 28S genes, multiple copies were found to be 164 dispersed throughout the genome, however, the majority, 97.1% of 18S, 97.6% of 5.8S, and 97.2% of 165 28S, were located in a single main 45S rDNA locus at the end of chromosome 18 (Fig. S3 A, B). The 5S 166 genes occur in three distinct contexts: (i) within the 45S locus forming 18S –5.8S–28S–5S arrays (155 167 copies), (ii) as 20 interspersed single copies scattered across the genome, and (iii) in two large tandem 168 arrays on chromosomes 13 and 21 (Fig. S3 A, C), which contain 151 and 208 paralogues, respectively. 169 These data expand our knowledge of 45S and 5S rDNAs in H. antarcticus , which were previously 170 thought to be co -localised to a single locus [26]. However, the dispersed nature of the 5S locus is 171 consistent with reports for other notothenioids and fish in general [26]. 172 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 7 Haplotypic variation in the H. antarcticus genome 173 Alignment of the two haplotypes of the fHarAnt.1. 2 assembly showed heterozygosity rate of 0.15%, 174 consistent with the low heterozygosity estimated based on k-mers, and a generally low number of 175 structural variants ( Table S 6; Fig. S4). More specifically 1.68% of the genome was present in a 176 hemizygous state (i.e., composed of heterozygous insertions and deletions), and 8.9 Mb and 9.1 Mb 177 was affected by heterozygous inversions on Hap1 and Hap2, respectively. Hemizygous deletions 178 relative to Hap1 affect a total of 849 exons, and 136 genes are completely deleted on Hap2 and thus 179 might be subjected to presence-absence variation at the population level. Importantly, 66.6% of all SVs 180 were also detected in a heterozygous state when performing a read-based SV call, while only 1.7% were 181 genotyped as homozygous for the alternate allele, indicating low levels of false positives due to 182 assembly errors. The remaining 31.7% were not detected using ONT reads. Using generalized linear 183 mixed models and chromosome identity as random effect, a significant positive association was 184 observed between SV occurrence (considering inversions, duplications, and translocations) and 185 repetitive content within 50 kb flanking regions of the breakpoints. The strongest predictor of SV 186 occurrence was the density of tandem repeats, with an odd ration of 2.7. Additionally, we found a 187 negative association of SV occurrence with gene content, distance from both chromosome ends , and 188 assembly gaps as well as LINE-L2 content (Fig. 3A, B). 189 The phased assembly allowed the inspection of the genome for the detection of potentially recently 190 active TEs , as recent insertions are expected to be found mostly in a heterozygous state and thus 191 correspond to insertion/deletion events between the two haplotypes ( Fig. 3C). Overall, 20% of the 192 insertions/deletions have a reciprocal overlap of at least 75% with annotat ed TEs and might therefore 193 correspond to heterozygous TE insertions. Accordingly, these TEs have a significantly lower 194 divergence from their consensus sequences than other TEs, which is used as a proxy for insertion time 195 (One-tail Wilcoxon rank sum test, p-value <0.001) (Fig. 3D). A 39.1% of all heterozygous TE insertions 196 correspond to LINE -L2 elements followed by unknown LTRs with 17.2%, and unknown DNA 197 transposons with 12.4% (Fig. 3E; Fig. S5). Among all heterozygous TE insertions, most are located in 198 intergenic regions (9,578), followed by 2,500 bp gene flanking regions (3,800) and introns (1,958). 199 Only 23 insertions affect coding exons, leading to allelic pseudogenization of the impacted genes. 200 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 8 201 Fig. 3: Haplotypic variation in the H. antarcticus genome. A. Detail of the right end of chromosome 202 15 harbouring multiple structural variants (SVs). B. Association between different genomic features 203 and SV distribution, using a randomized control expressed as log -odds ratios. Points represent the 204 estimated coefficients from the mixed -effects logistic regression. Horizontal lines indicate 95% 205 confidence intervals (“Dist. Chr . end”: distance from closest chromosome end “Dist. Gaps ” distance 206 from closest genomic gap percentage (%) of TEs, LINEs, genes, and tandem repeats shown along 50 207 kb genomic regions flanking the SV breakpoints, SV: length of SVs). C. Schematic representation of a 208 heterozygous LINE insertion between haplotypes. The blue rectangle marks the reverse transcriptase 209 domain. D. Percentage of divergence of TEs from their source consensus sequences as a proxy of the 210 time of the insertion of heterozygous, and all other TEs. E. Contribution of TE groups to heterozygous 211 insertions (see detailed figure in Fig. S5) (*** = p-value <0.001). 212 Genome evolution during notothenioid diversification 213 Synteny analyses of 12 notothenioid chromosome-scale genomes (Fig. 4A; Table S5) revealed strong 214 conservation of macrosyntenic relationships during their diversification, together with lineage-specific 215 Robertsonian chromosomal fusions ( Fig. 4B ), in line with previous karyotype analyses [27]. Most 216 assemblies, including H. antarcticus, show the predominant notothenioid karyotype of n=24, which 217 largely comprises acrocentric chromosomes with the exception of Notothenia rossii (Nototheniinae) 218 (n=12), and Pagothenia borchgrevinki (Trematominae) and Pogonophryne albipinna 219 (Artedidraconidae) (n=23). In N. rossi 10 of the chromosomes appear to have originated from the 220 Robertsonian fusion of 10 ancestral acrocentric pairs, one from the fusion of three chromosomes, and 221 one remained unfused leading to a karyotype of n=12. In P. borchgrevinki and P. albipinna, only two 222 different chromosomes have been fused, resulting in n=23 karyotype in both species. Across all 223 notothenioid genomes , detailed molecular mapping detected only a few large inversions and 224 intrachromosomal translocations were detected, mostly at chromosome ends (Fig. 4B; Fig. S4). 225 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 9 We used a whole genome alignment of 12 chromosomal assemblies to understand patterns of DNA gain 226 and loss during notothenioid diversification, and quantified the amount of bp involved in insertion and 227 deletion events across all internal and terminal branches. Overall, genomic gain greatly outweighs 228 genomic loss which accelerates towards recent times, reaching a maximum net rate of 76 Mb per MY 229 along the ancestor of the Channichthyidae family (Fig. 4A; Table S7). Along the cryonotothenioid stem 230 branch we observed a net genome expansion of 256 Mb. Although this represents the fourth largest 231 absolute gain across all branches, surpassed only by the terminal expansions in N. rossii, H. antarcticus, 232 and P. borchgrevinki, the long branching time resulted in a relatively modest rate of genome gain of 233 16.46 Mb per MY (Fig. 4A; Table S7). 234 Between 3,107 and 51,757 of the insertions inferred along H. antarcticus ancestors, map to an annotated 235 TE element of its genome, requiring a reciprocal overlap of 75% between the two annotations and thus 236 likely correspond to genuine TE insertions (Table S8). The number of TE insertions per MY greatly 237 increased during early evolution of cryonotothenioids, between 10.7 - 6.86 MYA, slightly decreased 238 along the lineage leading to the Harpagiferidae and Artedidraconidae sister species, and reached a main 239 spike of activity along H. antarcticus terminal branch (10,456 insertions per MY; Fig. 4C). DNA 240 transposons were the most active along all branches, but LINEs were predominant along the 241 cryonotothenioid stem branch (41% DNA vs 42% LINEs) and, particularly (Table S8), H. antarcticus 242 terminal branches (33% DNA vs 46% LINEs). Patterns of TE activity, especially for LINE elements, 243 mirror variations in temperature data, based on δ 18O variability [28] (Fig. 4C ), with higher values 244 related to lower temperatures. 245 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 10 246 Fig. 4: Genome evolutionary dynamics during notothenioid diversification. A. Dated phylogeny of 247 notothenioid species included in all comparative genomic analyses. Branch colours show the rate of net 248 genomic gain per million years as megabases per million years (Mb x MY). B. Syntenic relationships 249 among notothenioid chromosomes. Chromosomes were named based on synteny with C. gobio 250 assembly. C. Number of TE -related insertions along H. antarcticus ancestral and terminal branches 251 (highlighted in red in the phylogenetic tree above), shown as a smoothed curve. Scatterplot and black 252 curve show temperature variation over time based on deep-sea δ18O records (data from [28]). 253 Conserved elements and selection analyses on protein-coding genes 254 We identified genes and non -coding regulatory elements that may have facilitated the successful 255 colonization and diversification of cryonotothenioids in the cold waters of the Southern Ocean. First, 256 we analysed intergenic conserved elements (CEs) which are arising along their stem branch and are 257 conserved among all analysed species based on the notothenioid whole genome alignments, and second, 258 we investigated signs of diversifying selection on protein coding genes along the same ancestral branch 259 using a gene-wise branch-site model. We expect that genomic innovations underlying cold adaptation 260 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 11 arose along the notothenioid ancestral lineage. Although some cryonotothenioids such as D. eleginoides 261 and C. esox, inhabit more temperate regions, by colonizing Patagonia and several sub-Antarctic islands 262 [29,30], these are nevertheless considered secondary adaptations [31]. Prediction of CEs yielded a total 263 of 675,818 conserved genomic regions accounting for a total of 78,685,553 bp (Fig. S6A). Most CEs 264 (78%) are shared by all notothenioids, whereases only 10,182 emerged in the cryonotothenioid ancestor 265 and are shared by all members of the Antarctic clade (cryonotothenioid-specific CEs) (Fig. S6). Among 266 the cryonotothenioid-specific CEs, most correspond to intergenic (42.9%) and exonic (29.3%) genomic 267 regions. Intergenic, non-coding CEs (CNEs) might play a regulatory role in gene expression. We found 268 that the 1,261 genes associated with at least one cryonotothenioid -specific CNE were enriched for a 269 total of 236 Gene Ontology (GO) terms, most of which are related to signalling, cell differentiation and 270 developmental processes ( Table S9). However, we also identified 120 genes involved in biological 271 functions known to be important for cold adaptation, in processes such as antioxidant activity, lipid 272 metabolism, protein folding, and cytoskeleton organization ( Table S 10). Some genes of particular 273 interest including prdx5 associated with antioxidant activity (Fig. S6B), ube2e1 and ube2h, involved in 274 protein ubiquitination; ahsa1 which has chaperone -binding activity, and tubb4a, with Antarctic fish 275 tubulins previously having been identified as cold adapted at the amino acid level [32]. 276 For selection analyses we used an extended dataset which includes an additional seven temperate or 277 tropical fishes tagging on the cryonotothenioid stem branch ( Fig. S7). To avoid conflating the signal 278 with lineage -specific evolutionary dynamics, we did not test for diversifying selection in any 279 cryonotothenioid child branch. Furthermore, to establish a stronger link between the identified genes 280 and cold adaptation, we also excluded genes that were found to be under diversifying selection in any 281 other non-cryonotothenioid temperate or tropical fish, based on a control analysis (Fig. S7). While we 282 acknowledge that this latest approach might increase the false negative rates, excluding genes that might 283 be related to both cold -adaptation and other phenotypic traits due to for example substitution on 284 different codons [33] we prefer to automatically exclude genes that might be subjected to pervasive 285 diversifying selection along multiple branches of the phylogeny. We identified a total of 334 genes 286 under selection on the cryonotothenioid ancestor but not in temperate -adapted species, of which 256 287 were functionally annotated (Table S11). A total of 54 GO terms were enriched among the se genes 288 (Table S12). Of these, several are potentially linked to cold adaptation, particularly to oxidative stress 289 (GO:0070301, GO:0010039 ), lipid metabolism ( GO:0006084, GO:1903727, GO:0006690 ) and 290 proteostasis (GO:0051260, GO:0051603, GO:0006509, GO:1901984 ). Example gene include the 291 sirtuin 1 ( sirt1), peroxiredoxin 3 ( prdx3), mitochondrial superoxide dismutase 2 ( sod2), and 292 dihydrolipoamide dehydrogenase (dld) genes. 293 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 12 Characterization and evolution of the antifreeze gene locus 294 The antifreeze gene locus was located on chromosome 16, spanning approximately 699 kb on each 295 haplotype, in a region flanked by the hsl and tomm40 genes. Using a combination of automated and 296 manual approaches, we identified 17 and 16 afgp gene copies on Hap1 and Hap2, respectively, as well 297 as two chimeric afgp/tlp genes on each haplotype (Fig. 5A). The gene annotation completely matched 298 the afgp-specific k-mer signature identified by Klumpy suggesting a complete representation of the 299 locus (Fig. S8). Repeat coverage of the afgp locus was 58.7% (Hap1) and 59.1% (Hap2), approximately 300 1.43 times higher than the genome-wide estimation (Fig. 5A). 301 The typical structure of a functional afgp gene encodes for an exon -1 signal peptide and an exon -2 302 AFGP polyprotein with XLF spacer and AARG terminal motif [4] (Fig. 5B). On Hap1, 14 of the afgp 303 genes identified contained the correct structure and were thus considered to be functional. All complete 304 afgp copies except one were automatically predicted by GeMoMa, highlighting its strong ability to 305 detect intact afgp genes (Fig. S8). Three more identified copies were putative pseudogenes (Table S13), 306 two of which miss the exon -l signal peptide ( afgp1 and afgp17) (Fig. 5A). The third copy ( afgp4) 307 encodes a partial AFGP polyprotein with an unrelated amino acid tail sequence that lacks the canonical 308 AARG terminal motif. The two chimeric afgp/tlp genes contain a signal peptide, but while afgp/tlp1 309 encodes the XLF spacer and a short TLP segment, afgp/tlp2 encodes an unrelated amino acid tail 310 sequence and lacks the XLF spacer, making it a putative pseudogene (Fig. S9). 311 The structure of the afgp locus of fHarAnt1.2 is almost identical in both haplotypes, without evidence 312 of major rearrangements. One afgp gene copy is missing in Hap2 (afgp6) compared to Hap1, but this 313 may be likely due to a mis -assembly resulting in a gap ( Fig. 5A). Only insertions and deletions were 314 identified between the two haplotypes, with 48 short variants (<50bp), involving a total of 428 bp and 315 other 16 with a length from 51 bp to 7,477 bp affecting a total of 32,109 bp. The previously mentioned 316 pseudogenization of afgp4 on Hap1 resulted from a single -base deletion that shifts the reading frame 317 relative to Hap2, which retains a functional copy. Here, 22 insertions/deletions occur on exon2 of afgp 318 or afgp/tlp genes affecting a total of 3,27 k b. These variants do not disrupt the reading frame but 319 generate length allelic variants of the AFGP polyprotein. Specifically, we identified five large indels 320 with a length between 51 bp and 1,224 bp affecting a large portion of exon-2 of afpg7, afgp9 and afgp11 321 genes, respectively (Fig. 5B). Finally, the chimeric afgp/tlp2 gene on Hap1 is affected by the largest 322 identified SV within the locus: a large deletion of 7,334 bp that involves all tlp-derived exons encoding 323 the TLP mature protein (Fig. 5B). 324 The locus is characterized by a clear and repetitive organization with 14 afgp gene copies found in an 325 area spanning 287 ,044 bp. The genes and TEs in this region are arranged in a characteristic motif 326 consisting of two pairs of afgp genes with alternating orientation, and a specific set of TE insertions 327 (Fig. 5Aii). The motif is consistently repeated along the region suggesting that it mainly consists of 328 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 13 segmental duplications (Fig. 5A). Mapping of ONT, and 10X reads (minimum mapping quality >0, and 329 excluding secondary alignments) from [6] which come from the same individual, revealed sharply 330 reduced coverage within the previously mentioned central region of the locus ( Fig. S10 A, B, C). On 331 average, coverage decreased by 16.8% for ONT reads and by 55.5% for 10X reads. These values reach 332 the 85.4% and 32.8% when only reads with a minimum mapping quality of 20 are considered. Under 333 this threshold 32.9% of the locus show no coverage of 10X reads. 334 To better understand the evolutionary dynamics and duplication history of afgp and afgp/tlp genes we 335 inferred a maximum likelihood tree based on the conserved exon -1 and the starting region of intron -1 336 which is shared between afgp and chimeric genes ( Fig. S11; Fig. S12 ). Due to the difficulties in 337 resolving relationships among highly similar afgp gene copies, based on few alignment positions (391, 338 of which 56 are parsimony informative) ( Fig. S12A), we additionally performed a ML tree inference 339 excluding chimeric genes (Fig. S12B). This allowed us to include the full intron -1 and analyse a total 340 of 1,964 positions with 3.2 times more parsimony -informative sites. Because both trees recovered the 341 same two main afgp clades, we combined the two topologies by placing the chimeric genes onto the 342 afgp-only tree, preserving their phylogenetic positions ( Fig. 5C). The chimeric afgp/tlp2 genes from 343 both haplotypes form a clade separated by a long branch from all other genes (Fig. S12A). Considering 344 that afgps likely originated from a chimeric gene (Chen et al., 1997), afgp/tlp2 probably represents the 345 most ancestral gene and was therefore used to root the tree. All other analysed genes form two clades 346 that perfectly separate genes on the plus strand from those on the minus strand, including the chimeric 347 afgp/tlp1, which is also the first to diverge within its clade. 348 To have a broader overview of the locus across cryonotothenioids we also annotated the locus in 349 additional species. We used an automated re -annotation process of the afgp locus by employing 350 GeMoMa and Klumpy, followed by visual inspection of candidate genes. To benchmark this approach 351 we used the afgp locus of species P. georgianus which had previously been manually annotated, which 352 we then re -annotated through the automated process. The automated approach identified the same 353 number of chimeric afgp/tlp (1) and afgp (15) genes in P. georgianus, organized in the same structure 354 as described by [6]. We then auto-annotated two additional species: P. albipinna and N. rossii (Fig. 5D; 355 Fig. S8). In P. albipinna the locus was distributed across different unlocalized scaffolds despite the use 356 of long-read sequencing technologies and Hi -C data (Fig. S8). In N. rossii we detected two chimeric 357 genes and 10 afgp genes. We performed comparison of the structure of the locus along a number of 358 species including our new and published annotations for other included species (Fig. 5D). Overall we 359 detected some notable similarities between species C. esox, C. gunnari, H. antarcticus , and N. rossii. 360 Specifically, in all these species the majority of afgp genes are arranged in arrays with alternating 361 orientation. However, each pair of afgp genes lack s the previously described TE motif that was 362 identified in the H. antarcticus afgp locus. 363 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 14 364 Figure 5: Structure and evolution of the antifreeze locus. A . Structure of the antifreeze gene locus 365 on both H. antarcticus haplotypes. Triangles represent different genes , and rectangles represent 366 transposable elements. Afgp genes are named from 1 to 16 based on their position on the locus. Asterisks 367 highlight pseudogenes and double-slash lines indicate assembly gaps. The grey panel (ii) on the right 368 corner shows a zoomed in version of the tandemly repeated genomic unit marked by the characteristic 369 TE motif. B. Top: simplified representation of afgp and chimeric afgp/tlp genes (exon and intron lengths 370 are not proportional to their actual size). Bottom: representative afgp and afgp/tlp chimeric genes 371 subjected to insertion/deletion events between the two haplotypes. Coverage refers to ONT reads 372 mapped back to assembly. C. Phylogenetic tree of afgp and chimeric afgp/tlp gene copies. Different 373 colours correspond to strand (+/ -). D. Structure of the afgp locus across notothenioids. All genes, 374 including pseudogeni sed copies are reported. Locus for P. borchgrevinki , C. esox , C. gunnari, C. 375 aceratus, P. georgianus and D. mawsoni is based on published work [6,8,9,34]. P. albipinna could not 376 be included in this analysis due to the locus being highly fragmented in the corresponding assembly 377 (Fig. S8). 378 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 15

379 Haplotype-resolved assembly of the Antarctic spiny plunderfish genome reveals low 380 haplotypic structural variation 381 Here, we present the first haplotype -resolved genome assembly of a notothenioid species and one of 382 the most contiguous notothenioid genome assemblies to date. The assembly was generated with ONT 383 reads which supported assembly more successfully compared to sequencing attempts with HiFi data. 384 The two haplotypes exhibit low levels of divergence and a relatively small number of structural 385 variations, which is consistent with the low k-mer based and SNP based heterozygosity (0.37% and 386 0.15%, respectively) of the H. antarcticus genome (Table S6; Fig. S4). Few exons and genes were 387 found to be absent from one of the two haplotypes, in contrast to findings in other metazoans, where 388 highly divergent and hemizygous genomic regions can represent more than 50% of a diploid genome 389 [35,36]. 390 Complex SVs, such as translocations, duplications, and inversions were found to colocalize towards the 391 repeat-rich and gene -poor chromosome ends ( Fig. 2A, Fig. S4). This was even more so in terms of 392 tandem repeat content, consistent with a model of elevated mutation rates in sub -telomeric genomic 393 regions [37]. Accumulation of repeats towards chromosome ends was also observed in another 394 cryonotothenioid species, T. borchgrevinki [34], so we suggest that this might be a plesiomorphic 395 feature of the clade. Furthermore, recent TE insertions were found to significantly contribute to the 396 emergence of insertions and deletions between haplotypes. Even though heterozygous TE insertions are 397 usually found in intergenic genomic regions likely due to strong negative selection against TE insertion 398 within genes, as observed in zebrafish [38], here a high number of TE insertions were detected in gene-399 flanking regions. In few cases, insertions within the exons of protein-coding genes were detected, which 400 have led to the truncation and putative pseudogenization of the affected genes on one of the haplotypes. 401 Overall, our results suggest that despite the low levels of overall haplotypic structural varia bility, 402 recently active TEs likely represent important contributors in driving population-level differentiation in 403 H. antarcticus, promoting gene-loss and reshaping cis- and trans- regulatory elements. 404 Furthermore, haplotype resolved assemblies would also have the potential to shed new light on overall 405 genome evolutionary dynamics, facilitating assembly -based variant calling [39], pangenome 406 reconstruction [40], allele-specific expression analyses [41], and TE identification [42]. 407 Recent accumulation of ancient LINE-L2 elements in H. antarcticus 408 Different TE clades can have varying evolutionary trajectories across species, reflecting host–element 409 co-evolution, lineage-specific regulatory constraints, and variation in genomic defence mechanisms 410 [43,44]. LINE-L2 transposons were previously shown to be particularly abundant in H. antarcticus 411 compared to other notothenioids, covering about 8% of the genome [6]. Analysis of the new genome 412 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 16 assembly similarly showed high LINE -L2 levels deriving mainly from the activity of eight different 413 LINE-L2 families. Both the high number of heterozygous LINE -L2 insertions and the repeat -414 divergence landscape analyses suggest a recent, lineage-specific accumulation of these elements (Fig. 415 2). Eight evolutionary distinct families of LINE-L2 elements were identified, with concurrent activity 416 within the last four million years (MY) (Fig. 2D). Moreover, most of these LINE-L2 families, including 417 the most abundant ones, are also shared with sister species of the cryonotothenioid clade, suggesting 418 their ancient emergence. Despite the recent activity of LINE-L2 elements, we observe a negative effect 419 of their genomic content around breakpoint regions and occurrence of inversions, duplications and 420 translocations between haplotypes suggesting that their accumulation has not resulted in the emergence 421 of other more complex structural variants. 422 In another LINE transposon family, the LINE-L1 retrotransposons, it has been shown that either a single 423 lineage was usually active at high replication rate at any given time in mammals, or multiple, ancient 424 LINE-L1 lineages were concurrently active in fishes, reptiles, and bivalves [44,45]. The evolutionary 425 path of LINE-L2s in notothenioids appear therefore to follow the generally observed evolution of LINE-426 L1 of fishes. 427 Rampant genomic gain and low levels of genomic deletion during cryonotothenioid 428 diversification 429 TE-mediated genome expansion has been shown to predate the diversification of extant 430 cryonotothenioids and that enhanced TE activity continued in more derived clades such as icefishes 431 [6,34]. To further test this hypothesis, we have quantified genomic gains through TEs along the 432 notothenioid ancestral branches ( Fig. 4A ). Mapped TE insertions along H. antarcticus ancestral 433 branches as well as a dated repeat landscape plot, indicate that most TE activity occurred recently, 434 concurrently with the recent cryonotothenioid diversification. A single massive burst of TE activity was 435 also detected in repeat profiles of other cryonotothenioids [6,15,34]. Low levels of ancient TE activity 436 followed by recent peaks could reflect either genuinely increased transposition rates or extensive 437 genomic deletions that removed older insertions, thereby erasing evidence of ancient bursts. However, 438 net genome -gain estimates reveal low deletion rates during notothenioid evolution, supporting the 439 former scenario: a slow accumulation of TEs before the emergence of extant cryonotothenioids, 440 followed by a more recent increased rate of TE accumulation . Our reconstruction also revealed an 441 ancient burst of genome expansion in the ancestor of icefishes explaining their exceptionally large 442 genomes, which may be explained by sea temperature history in the region [6]. 443 Paleoclimatic based temperature calculations ( Fig. 4C ) indicate an initial temperature drop at 444 approximately 16 MYA which predates the emergence time of the extant cryonotothenioid clade, and 445 a second, stronger reduction around 3 MYA which could have contributed to their recent diversification 446 [6], similarly to other marine clades [46]. Here we show that the timing of these temperature variations 447 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 17 appear to coincide with our reconstructions of TE levels of activity (Fig. 4C). LINE elements were 448 found to be particularly expanded not only in H. antarcticus but also in other cryonotothenioids (Fig. 449 2E). The presence of environment induced stressors , such as fluctuations of temperature , have been 450 linked to increased TE activity due to likely disrupting mechanisms of epigenetic regulation [47]. In the 451 cryonotothenioid species D. mawsoni, it has been shown that the levels of LINE element transposition 452 are increased through cold induced cell transfection [48]. We provide evidence that compared to other 453 TEs, LINEs have been particularly active along the cryonotothenioid ancestor and starting from 4 MYA, 454 concurrently with the two main temperature drops documented in the region (Fig. 4C). Overall, our 455 comparative genomics analysis suggests that the increased genome size of cryonotothenioids may have 456 occurred through a two -step process: one ancient, slow TE activity in the absence of major genomic 457 deletions that predate the emergence of extant species and one more recent, lineage -specific massive 458 bursts of transposition due to loss of epigenetic control. 459 Macrosyntenic relationships and chromosomal fusions in notothenioid genomes 460 Our synteny analyses of 12 chromosome -scale notothenioid genomes demonstrated conserved 461 marcosyntenic relationships between notothenioid chromosomes in accordance with the presence of a 462 stable karyotype with low structural divergence during notothenioid radiation [49]. In the Nototheniidae, 463 the most common chromosome number is 2n=48, mainly comprising acrocentric chromosomes. Indeed 464 H. antarcticus has a karyotype of 2n=48 comprising 42 acrocentric, 2 metacentric and 4 submetacentric 465 chromosomes [49]. Lineage-specific Robertsonian chromosomal fusions have occurred in some 466 notothenioids [50], with the most notable case being N. rossii where only 12 chromosomes (2n=24) are 467 observed (Fig. 4B). The few other identified smaller scale rearrangements tended to colocalize towards 468 repeat-rich chromosome ends, similar to haplotypic structural variants identified in the H. antarcticus 469 genome. These genomic regions are therefore thought to be subjected to higher rates of structural 470 variation compared to the rest of the chromosomes, with repetitive elements that may drive this 471 variability. Furthermore, the existence of LTR retrotransposon –insertion hotspots near chromosome 472 ends and at chromosome fusion points, suggests that such repetitive sequences may have facilitated 473 Robertsonian fusions, as has occurred in other more distant species [16,51]. In cryonotothenioids, 474 reduction in chromosome number from the ancestral karyotype of n=24, caused by multiple independent 475 chromosomal fusions, appears to be common within specific subfamilies, with such cases observed in 476 the genera Trematomus, Notothenia, and Pogonophryne [24]. 477 Tandem segmental duplications and haplotype length variants shape the structure of the 478 antifreeze locus 479 The antifreeze locus is a very challenging genomic region to assemble both because of its repetitive 480 structure due to the presence of tandemly duplicated gene families such as the afgps, and trypsinogens, 481 as well as due to its high TE content [6]. Using ONT reads we were able to assemble and annotate the 482 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 18 complete afgp locus for H. antarcticus for both distinct haplotypes. The locus comprises 17 afgp copies 483 (Hap1) and a higher repeat content compared to the genome average (59% vs 43%). One copy of the 484 chimeric afgp/tlp gene was identified on each haplotype (Fig. 5A), both of which are characterized by 485 canonical structure in accordance to previously described genes (Chen et al., 1997; Cheng and Chen, 486 1999; Nicodemus-Johnson et al., 2011) and are considered to be functional copies. These chimeric 487 genes were previously considered to be putative evolutionary intermediates of the afgp genes [7]. 488 Annotation of the locus on members of the icefish family showed either lack ( C. gunnari and C. 489 aceratus) or pseudogenisation ( P. georgianus ) of chimeric gene copies, suggesting that chimeric 490 afgp/tlp genes might have been lost in low latitudes of the Southern Ocean by relaxed selection due to 491 redundancy function with afgp genes [6,9]. However, chimeric genes have been retained in H. 492 antarcticus which has a partially overlapping distribution range with these species, leading us to 493 question the drivers of retaining these genes for this species. 494 BAC library -based analyses of the afgp locus of D. mawsoni suggested the presence of massive 495 variation between haplotypes in both the overall size of the locus, and in the number of afgp gene copies 496 [8]. Tandem arrays are generally expected to be frequently subjected to copy number variation due to 497 recombination between homologue chromosomal regions which could also promote high levels of 498 haplotypic variation [52]. Through this haplotype resolved assembly for H. antarcticus we were able to 499 directly investigate the potential presence of haplotypic divergence in the locus. Nevertheless, we found 500 that in H. antarcticus, both haplotypes exhibit almost identical structure and gene copy number. The 501 main structural variations identified where pseudogenizations of one afgp gene due to a single base-pair 502 deletion and of the only putatively functional afgp/tlp gene due to large deletion of all tlp derived exons 503 on Hap1. 504 Despite low levels of variation at gene copy and functionality levels, we also identified allelic variation 505 of the afgp gene copies, with regards to the actual length o f the AFGP polyprotein coding exons. 506 Potential variation on the length of the AFGP polyprotein could have important biological implications. 507 The notothenioid AFGP proteins function th rough an “adsorption –inhibition” mechanism [53], 508 adsorbing to the surface of ice and preventing the growth of the crystals as a function of glycoprotein 509 concentration, size, and shape [54,55]. It has been observed that smaller protein molecules are less 510 effective than large ones to interfere with ice crystal growth [54]. We could hypothesise that allelic 511 variation linked to the size of AFGP polyproteins in H. antarcticus populations could represent an 512 important source of genomic variation that allows local adaptation to more extreme conditions. 513 Length allelic variants have also been observed in other highly repetitive gene families, such as silk 514 genes in caddisflies, butterflies, and spiders [56], where sequence length was linked to the properties of 515 the silk fibres [57]. Such types of allelic variants may represent a convergence source of variation in 516 proteins composed of repeated motifs across all metazoans. In these cases we could consider that 517 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 19 because the gene itself provides a substrate for recombination, and the repetitive structure can weaken 518 purifying selection against insertions and deletions, these could create a standing pool of genetic 519 variation on which positive selection can eventually act. 520 In the case of the afgp locus, t he characteristic TE plus afgp gene motif that was detected in H. 521 antarcticus likely supports a specific mode of expansion of the locus. The afgp copies are arranged in 522 opposite orientation separated into two strand -specific clusters ( Fig. 5A, C), suggesting that the 523 expansion of the genes occurred through tandem segmental duplications, which duplicated as two pairs 524 of genes at a time. Furthermore, the high sequence homology identity across duplicated regions and the 525 short branch lengths separating gene copies in phylogenetic analyses (Fig. 5, Fig. S12) suggest that 526 these duplications are either evolutionarily recent or have been strongly homogenized by concerted 527 evolution. 528 After their emergence in the cryonotothenioid ancestor, reduction and loss of afpg genes has occurred 529 multiple independent times during their radiation [58]. However, the exact time at which afgp genes 530 begun duplicating and expanding remains an open question. Comparison of the afgp locus structure 531 across multiple notothenioid genomes (Fig. 5D ) showed a similar genomic configuration of some 532 genomic features in P. borchgrevinki, H. antarcticus, C. esox, C. gunnari, and N. rossii. Nevertheless, 533 the distinct TE motif appears to be specific to the H. antarcticus genome (Fig. 5Aii). We hypothesize 534 that the tandemly duplicated genomic region marked by the characteristic TE motif of H. antarcticus is 535 a plesiomorphic feature of the other species, with different rates of duplication and loss driving variation 536 of afgp gene copy number. We therefore suggest that from a hypothetical cryonotothenioid ancestral 537 locus with few afgp genes, independent tandem duplications increased gene-copy number in different 538 lineages. 539 Diversifying selection and origin of novel conserved elements in cold -adaptive pathways 540 during cryonotothenioid emergence 541 To detect additional genomic features underlying cold adaptation in notothenioids we searched for both 542 protein-coding genes that experienced diversifying selection and CNEs that first emerged along their 543 ancestral branch. Among the genes showing evidence of diversifying selection in the cryonotothenioid 544 ancestor, many were found to be associated with biological processes known to be important for 545 survival in extreme cold conditions (Table S12). We found that most CNEs were shared across all 546 notothenioid species, suggesting that a large portion of the genome expansion characterizing the 547 cryonotothenioid ancestral lineage may have evolved neutrally. Additionally, most cryonotothenioid-548 specific CNEs were found to be associated with genes involved in signalling, cell differentiation, and 549 developmental processes (Table S9). CNEs are known to have non -random distribution within the 550 genome, because they have the tendency to cluster near genes with regulatory functions in multicellular 551 development and differentiation (Polychronopoulos et al., 2017). A subset of the genes which we have 552 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 20 found to contain newly acquired CNEs are related to two biological processes which are 553 overrepresented among genes under diversifying selection. These biological processes are antioxidant 554 activity and proteostasis and they have been linked to evolutionary cold adaptation [59,60]. We suggest 555 that because these genomic innovations arose in the cryonotothenioid ancestor, they are amongst the 556 key components of these pathways that enabled this clade to survive and become established in the 557 Southern Ocean. 558 Antioxidant activity and proteostasis 559 Life in freezing sea water poses particular biological problems for the organisms living there. The 560 amount of oxygen increases considerably with reduced sea water temperatures, producing naturally 561 high levels of cellular oxidative stress for Antarctic marine species [59,61]. In addition, cold 562 denaturation of proteins impacts proteostasis [60]. Therefore, in order to survive in this harsh 563 environment, organisms need to modify their biochemical processes. 564 Our analysis revealed four key genes related to oxidative stress that underwent diversifying selection in 565 the cryonotothenioid ancestral lineage, including the sirtuin 1 ( sirt1), peroxiredoxin 3 ( prdx3), 566 mitochondrial superoxide dismutase 2 (sod2), and dihydrolipoamide dehydrogenase ( dld) genes. 567 Moreover, we found that the peroxiredoxin 5 (prdx5) gene acquired novel CNEs concurrently with 568 extant cryonotothenioid emergence. Gene sirt1 is a class III histone deacetylase that is recruited in the 569 promoter region of multiple antioxidant genes, including sod2, prdx3 , and prdx5, promoting their 570 expression and their protein concentration [62]. These three genes encode mitochondrial enzymes that 571 act in concert to control reactive oxygen species (ROS) levels. Sod2 converts superoxide radicals into 572 hydrogen peroxide, which is then detoxified to water by prdx3 and prdx5 [63]. The interplay between 573 protein modifications and the emergence of novel, potentially regulatory CNEs within this pathway 574 may have strengthened the antioxidant capacity of the cryonotothenioid ancestor, enabling it to cope 575 with elevated oxidative stress in the mitochondria. 576 Signs of diversifying selection have been found [64] also in other cold-adapted notothenioids acting on 577 genes involved in antioxidant activity, such as in N. coriiceps [65] and D. mawsoni [15], and across 578 species. Moreover, genes such sod2 and prdx5 which are involved in ROS scavenging, have been found 579 to be overexpressed in cold-adapted notothenioids compared to their temperate-adapted relatives [66]. 580 Superoxide dismutase appears to be particularly important in combatting oxidative stress in Antarctic 581 species. Gene duplications used to enhance the cellular activity of this enzyme have been identified in 582 Antarctic invertebrates [67]. Furthermore, evolution to function efficiently in the cold have resulted in 583 at least some members of this gene family being particularly thermally sensitive in some Antarctic 584 marine invertebrates [68]. This would suggest that specific protein modifications might help to mitigate 585 cellular oxidative stress in Antarctic species. 586 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 21 In terms of proteostasis, there is evidence that Antarctic species, including fish, face challenges in 587 producing, folding and maintaining proteins at such low temperatures [60]. Whole animal RNA to 588 protein ratios , ubiquitylation rates , and levels of chaperone proteins are much higher in Antarctic 589 compared to temperate species [69–71]. Some of this activity is due to a mutation in the promotor region 590 of the classical inducible form of hsp70 resulting in constitutive expression and cold-enhanced activity 591 of the proteosome in Antarctic notothenioids [72,73]. In our analyses on genes that were found to be 592 under diversifying selection, we identified genes involved in protein folding and degradation such as 593 members of heat shock protein family member (hspa4), ubiquilin/ubiquitin proteins (ubqln4 and ube2z) 594 and the chaperon psmc1. Similar to genes with acquired CNEs, such as the ubiquitin ube2h, ube2e1, 595 itch and the co -chaperon ahsa1. Furthermore, there was evidence of selection pressures on genes 596 involved in transcription, splicing, and the proteasome (e.g. eif3b, sf3a2 and psmb10) indicating wider 597 scale cold adaptation of the proteostasis pathway. 598 It should be noted that these examples are likely minimal identifications due to the stringency of 599 orthologue assignment, which may not accurately identify duplicated genes, the number of which are 600 considerable in the notothenioids [66]. Such analyses will also fail to identify those genes which are 601 poorly conservated between species, such as genes involved in the immune function, intrinsically 602 disordered proteins and those containing intrinsically disordered regions. 603

604 This haplotype-resolved genome assembly of H. antarcticus thus provides a valuable resource for future 605 functional investigations into cold adaptation and resilience to climate change. Overall haplotype 606 diversity was low and contrary to expectations haplotypic diversity in the afgp locus was also low 607 despite a high number of transposon insertions. However, we did identify sequence length variation in 608 the afgp genes between the two haplotypes, which could be biologically relevant. Furthermore, we show 609 that genomic gain was linked to transposon activity which correlates with drops in temperature in the 610 Southern Ocean, and this activity resulted in the generation of novel genomic elements . Finally, we 611 show that there has been significant selection pressure on notothenioid genomes, both in genes and 612 control regions, to enable these fish to thrive for much of their life below 0°C. 613 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 22

614 Sequencing data generation 615 The specimen of Harpagifer antarcticus that was used for genome assembly was collected at Ryder 616 Bay in the Antarctic peninsula (coordinates: -67,59053 -68,28863), and was preserved by flash freezing. 617 This specimen was the same that was used for the generation of the fHarAnt1 .1 assembly (accession: 618 GCA_902827135.1; BioSample: SAMEA104132831) [6]. 619 High Molecular Weight (HMW) DNA was extracted using the Sanger Tree of Life Manual MagAttract 620 v1 protocol [74] in the Tree of Life Core Laboratory at the Wellcome Sanger Institute. To assess the 621 quantity of the extracted DNA the QubitTM 1X dsDNA High Sensitivity assay (ThermoFisher, Cat No. 622 Q33231) on the Qubit Flex Fluorometer (ThermoFisher, Cat No. Q33327) was used. The quality of the 623 DNA was then determined using the Femto Pulse (Agilent, PN: M5330AA) instrument with the gDNA 624 165 kb Analysis kit 275 samples (Agilent, Cat No. FP-1002-0275). Approximately 3µg of the extracted 625 DNA underwent library preparation for ONT sequencing using the Ligation Sequencing Kit V14 (SQK-626 LSK114). Sequencing was performed on one flow cell on the PromethION 24 in Sequencing Operations 627 at the Wellcome Sanger Institute. 628 Three H. antarcticus specimens were Schedule 1 killed and the gills rapidly dissected, added to tissue 629 culture media (L15 without phenol red, 10% foetal bovine serum and antibiotics) and placed on ice. 630 The gills were chopped up in culture media and the resultant cell suspensions of each sample added to 631 individual wells in 6-well Nunclon™ tissue culture plates. For each fish, one of each of the gill samples 632 was incubated at 0°C with the paired sample incubated at 6°C for 6 days. Cells were harvested and 633 RNA extracted using TRI Reagent according to manufacturer’s instructions. 1µg of rRNA depleted 634 RNA was used for RNAseq library preps with Kapa Hyperprep UDI kit. (Cat#KK8544). Paired end 635 sequencing was carried out by Novogene on the Novaseq PE150 platform. 636 Genome assembly 637 The ONT reads were sequenced at an estimated 23.4X coverage per haplotype and based called with 638 dorado 7.2.13 (https://github.com/nanoporetech/dorado) in super accurate mode. The reads were 639 assembled using Hifiasm version 0.24.0 -r702 [75] with --ont option in Hi -C phasing mode. This 640 produced a phased assembly with contig N50 of 19.3 Mb and 14.8 Mb in each haplotype. The sizes of 641 both haplotypes were 1.072 Gb and 1.222 Gb which were close to the kmer -based estimate from the 642 GenomeScope at 1,152 Gb. 643 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 23 Furthermore Hi-C reads were mapped to each haplotype independently using bwa -mem2 [76]. Each 644 haplotype was then scaffolded with Hi -C data in YaHS [77], using the --break option for handling 645 potential mis-assemblies in contigs. The produced assembly was evaluated for kmer completeness using 646 MerquryFK (https://github.com/thegenemyers/MERQURY.FK) using kmers from the ONT reads. This 647 indicated 99.39% kmer completeness and QV value 42.6. 648 The individual haplotype assemblies were screened for contamination using the ASCC pipeline 649 (https://github.com/sanger-tol/ascc) removing 5.8Mb Hap1 and 1.8Mb Hap2 non -target and 650 mitochondrial sequences. The assemblies were then combined and prepared for curation using the 651 treeval pipeline ( https://github.com/sanger-tol/treeval) generating a Hi -C contact map with 652 supplementary analysis tracks (telomeres, gaps and long read coverage). Assembly errors were 653 corrected by manipulating the contact map in PretextView (https://github.com/sanger-tol/PretextView) 654 an AGP of the resolved chromosomal haplotypes were exported and a curated fasta for each haplotype 655 generated using pretext -to-asm (https://github.com/sanger-tol/agp-tpf-utils) with each super scaffold 656 named according to descending size. Contact maps for each resolved haplotype were generated by 657 curationpretext ( https://github.com/sanger-tol/curationpretext) and images generated with 658 PretextSnapshot (https://github.com/sanger-tol/PretextSnapshot). 659 Repeat annotation 660 Repetitive elements were de-novo mined with RepeatModeler v.2.0.5 [78] adding the -LTRStruct 661 extension to improve LTR detection. Potential multicopy host genes included in the repetitive library 662 were removed with ProtExcluder v.1.2 after blasting (Blastx, e -value 1e -10) the repetitive library 663 against a protein database consisting of all vertebrate proteins included in the Swiss-Prot database (The 664 UniProt Consortium, 2025) and the predicted proteomes of five published notothenioid genomes 665 available on NCBI RefSeq: C. gobio, P.georgianus, Gymnodraco acuticeps, Trematomus bernacchii, 666 Eleginops maclovinus . Furthermore, a TE library was made with MCHelper v.1.7.0 with default 667 parameters. Both haplotypes were annotated with RepeatMasker in sensitive mode (-s). RepeatMasker 668 derived annotation was post -processed with: (a) parseRM.pl script 669 (https://github.com/4ureliek/Parsing-RepeatMasker-Outputs/tree/master) to summarize the TE content 670 and obtain TE landscapes describing transposon activity through relative time, and (b) the RM2Bed.py 671 script from RepeatMasker to resolve overlaps in the TE annotation (--overlap resolution ‘higher score’) 672 and produce bed files describing the coordinates of TE insertions. 673 Tandem repeats were identified with the Centrominer function of quarTeT v.1.2.5 [79]. rDNA genes 674 were annotated with barrnap v.0.9 [80] https://github.com/tseemann/barrnap and telomeric repeats with 675 tidk v0.2.31 [81] screening the assembled chromosomes for the TTAGGG telomeric motif. 676 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 24 Gene annotation 677 Gene annotation of Hap1 was performed with BRAKER3 [82]. Repetitive elements were soft masked 678 and both protein and transcriptomic data were used as external evidence to train AUGUSTUS [83] and 679 GeneMark-ETP [84] ab-initio gene predictors. Illumina RNAseq reads from muscle, skin, liver, kidney, 680 heart, and brain [5] (Table S14) and from gills (data generated herein) were mapped with HISAT2 [85]. 681 The sorted bam files were supplied to BRAKER3 together with the protein database previously used 682 for cleaning the repetitive libraries from host genes. The quality of the annotation was assessed with 683 BUSCO v.5 [86]. To obtain a gene annotation for Hap2, we mapped BRAKER -derived genes with 684 Liftoff [87] (-polish -copies) to obtain their lifted coordinates. Functional annotation was performed on 685 the Hap1 predicted proteome with eggNOG -mapper v2 [88] transferring only annotations with 686 experimental evidence and auto-adjusting the taxonomic scope based on the query sequence. 687 Annotation of the afgp locus 688 To annotate the afgp locus on each haplotype, a combination of automated and manual approaches was 689 used. The annotation of the D. mawsoni afgp locus (accession HQ447059.1, haplotype 1) [8] was used 690 as a reference for homology-based gene prediction with GeMoMa v.1.7.1 [89] under default parameters. 691 This initial annotation was supplemented by manually blasting D. mawsoni gene sequences with 692 BLASTN v.2.12.0 [90] and reconstructing the gene structure based on alignment coordinates. Each 693 identified afgp copy was then manually inspected for frameshifts and gaps to identify complete and 694 pseudogenised gene copies. Finally, we used Klumpy [91] with the find_klumps function to verify the 695 completeness of the annotation, using the exon -1 and exon-2 of one randomly chosen functional afgp 696 gene copy previously annotated as query. 697 To infer evolutionary relationships of afgp genes, we used the short exon-1 (signal peptide) and intron-698 1 nucleotide sequences, excluding the highly repetitive exon -2s encoding the AFGP polyprotein, 699 following the Nicodemus -Johnson et al. , (2011) approach. Analysis was performed both with and 700 without the chimeric afgp/tlp genes. Sequences were aligned with MAFFT v.7.525 [92] using the E-701 INS-i. Because afgp and afgp/tlp genes share only exon -1 and an initial part of intron -1, when these 702 latest genes were included in the analyses, we removed the non -homologue region based on visual 703 inspection of the alignme nt. Phylogenetic inference was performed with IQ -TREE 2 [93] with 1,000 704 ultrafast bootstrap replicates [94], to assess nodal support, selecting the best-fit evolutionary model with 705 ModelFinder [95]. 706 Characterization of LINE-L2 elements 707 LINE-L2 insertions annotated on Hap1 and longer than 1,000 bp were extracted with bedtools getfasta 708 and subjected to Blastx ( -evalue 1e -05 –max_target_seqs 1 –max_hsps 1) against all reverse 709 transcriptase (RT) amino acid (aa) sequences that compose the seed alignment of the Reverse 710 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 25 transcriptase PFAM [96] profile PF00078. Based on the BLASTX v.2.12.0 [90] alignment coordinates, 711 the amino acid sequences of the RT segments were extracted, discarding those shorter than 100 amino 712 acids. To decrease the size of the dataset for the subsequent phylogenetic analyses all RT segments 713 were clustered at 80% sequence identity with CD -HIT [97] keeping the longest element as 714 representative of the cluster. All representative RT segments together with reference sequences of the 715 LINE-L2, L2A, L2b, Crack, Daphne and CR1 clades described in [25] were aligned with MAFFT in 716 G-INS-i mode and trimmed with Trimal v.1.4.1 [98] in gappyout mode. Phylogenetic inference was 717 performed with IQ-TREE v2.4.0 [93] with 1,000 UltraFast bootstrap replicates to assess nodal support 718 and selecting the best-fit evolutionary model with ModelFinder. 719 CD-HIT [97] clustering identified eight clusters, each comprising between 600 and 3,669 RTs, together 720 accounting for 73% of all extracted RTs longer than 100 aa. The complete insertion corresponding to 721 the representative element of each one of these clusters was used as query in a “Blast-Extend-Extract" 722 analyses [99–101] to reconstruct a consensus sequence representative of the family. The curated 723 consensi were used in an additional RepeatMasker analyses, as previously described. We finally looked 724 for the identified and reconstructed LINE -L2 family across genomes (Table S5) through BLASTN 725 searches requiring an alignment length of at least 500 nucleotides with at least 80% of identity to the 726 query consensus sequence. 727 Haplotype comparison 728 A whole genome alignment of the two haplotypes was performed with Minimap2 v.2.26 [102] and 729 inferred SNPs, SVs (insertions, deletions, and duplications), and rearranged genomic regions 730 (inversions and translocations) between the two haplotypes using Syri v.1.7.1 [103]. In each case Hap1 731 was used as the reference. We identified genes and exons subject to structural v ariation between the 732 two haplotypes by intersecting the gene annotation with Syri-derived SV and rearrangement coordinates 733 using bedtools v2.30 [104] intersect. We tested whether the genomic distribution of SVs, considering 734 only translocations, duplications and inversions, was associated with specific genomic features using a 735 generalized linear mixed model (GLMM) with a binomial error structure. SVs were compared to an 736 equal number of randomized genomic intervals generated with bedtools shuffle. For each interval we 737 quantified the following: distance to the nearest chromosome end, distance to the closest assembly gap, 738 length of the SV, and the proportion of bases within 50 kb flanking windows annotated as TEs, tandem 739 repeats, or LINE elements. All predictors were standardized prior to analysis. The model included 740 chromosome as a random intercept to account for non -independence of intervals on the same 741 chromosome. GLMMs were fitted in R using the lme4 package [105]. Odds ratios and 95% confidence 742 intervals were obtained by exponentiating model coefficients. 743 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 26 To identify putative heterozygous TE insertions, we overlapped the RepeatMasker annotations of Hap1 744 and Hap2 with deletions and insertion events, respectively, requiring that both insertions/deletions and 745 the TE annotation overlap by at least 75% of their length. 746 To assess whether the SVs identified between haplotypes might result from assembly errors, we 747 additionally mapped ONT reads back to the assembly with Minimap2 and called heterozygous SVs 748 with Sniffles2 [106]. We then quantified the overlap between the genome -based and read -based 749 heterozygous SVs larger than 50 bp using SURVIVOR [107]. Overlapping SVs genotyped as 750 homozygous for the alternate allele were considered as putative mis-assemblies. 751 Species tree reconstruction and divergence time estimation 752 Divergence times were estimated across a set of genomes of notothenioids, including our newly 753 generated H. antarcticus assembly (Hap1), and 11 publicly available chromosomal assemblies : 754 Cottoperca gobio (Cottoperca trigloides) [23], Notothenia rossii [108], Pseudochaenichthys 755 georgianus [6], Eleginops maclovinus [109], Dissostichus mawsoni [110], Dissostichus eleginoides, 756 Chaenocephalus aceratus [111], Pagothenia borchgrevinki (Trematomus borchgrevinki ) [112], 757 Champsocephalus gunnari, Champsocephalus esox [9], Pogonophryne albipinna [113] (Table S5). 758 Additionally another seven marine temperate and tropical fishes' representative of four, closely related 759 orders (Table S5). 760 First, we manually constructed a topology of the included species following the [6] and [9] inferred 761 phylogenetic relationships. Branch lengths were then estimated using the nucleotide sequences of 762 orthologous single -copy BUSCO genes extracted with BUSCO v.5.7.1 [86] and the 763 actinopterygii_odb10 reference dataset. Briefly, single-copy genes present in all analysed species, were 764 separately aligned using MAFFT [92] with the E -INS-i algorithm. Each alignment was cleaned of 765 ambiguous positions using TrimAl in automated mode, and sequences sharing only a reduced region 766 with the others in the alignment were removed using the parameters -resoverlap 0.8 and -seqoverlap 75. 767 Finally, only alignments with at least 500 positions and for which no sequences were excluded by 768 TrimAl were retained. To reduce computational time, 1,000 alignments were randomly selected, 769 concatenated, and subjected to constrained phylogenetic inference with IQ-TREE v2.4.0, selecting the 770 best-fit evolutionary model with ModelFinder. The resulting tree was dated with LSD2 [114] under a 771 fixed substitution rate using [6] estimates as priors of the following nodes:1) split between Takifugu 772 rubripes and Nelusetta ayraudi from all other species: 94 MYA, 2) notothenioid emergence: 47 MYA, 773 and 3) cryonotothenioid diversification: 10.7 MYA. A neutral substitution rate of notothenioids was 774 obtained by re-running the same analyses on the notothenioid-only subtree and using 4-fold degenerate 775 sites extracted from the BUSCO concatenated alignment. The resulting neutral rate of 3.2E -3 million 776 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 27 year was used to translate % of CpG corrected Kimura Divergence of TE landscape profiles into 777 absolute time. 778 Selection analyses 779 Publicly available predicted proteomes of fish species included in divergence time estimation were 780 clustered into orthogroups (OG) with Ortho Finder v2.5.4 [115] in ultra-sensitive mode. We then 781 isolated OGs in which at least one gene for all species was present. Multi-copy OGs (i.e. more than one 782 gene was present for at least one species) were decomposed into single-copy OGs using DISCO [116]. 783 Decomposed and original single-copy OGs were then aligned with Prank v250331 [117] under a codon 784 model, cleaned of ambiguous positions with Gblocks taking into consideration the codon structure [118] 785 and from sequences with high percentage of gappy positions with TrimAl (-resoverlap 0.6 -seqoverlap 786 55) [98]. After filtering, only OGs for which all species were present were kept (i.e. not affected by 787 Trimal filtering) and with an alignment longer than 300 codon positions. IQ-TREE2 v2.4.0 was used to 788 infer gene trees under a GTR+I+G substitution model. Because our aim was to detect genes that might 789 be strictly related to cold adaptation in extant cryonotothenioids, we used aBSREL from the HyPhy 790 package v2.5.8 [119] testing for diversifying selection acting on their stem branch. Prior to this, to 791 ensure that we were testing the same evolutionary event we discarded any gene trees in which 792 cryonotothenioids were not monophyletic and not in a sister relationship with their true temperate 793 notothenioid sister species E. maclovinus using a custom R script. Finally, because aBSREL analyses 794 does not test that selection has not occurred outside tested branches we re-ran it for all genes previously 795 identified as under diversifying selection along the cryonotothenioid stem tagging all temperate-adapted 796 species (including non-cryonotothenioid notothenioids) and all their ancestral branches. OGs that were 797 found to be under positive selection in at least one temperate-adapted branch were then removed from 798 aBSREL significant results on the cryonotothenioid ancestor. GO term enrichment was performed using 799 all genes analysed by aBSREL as background based on EGG -NOG annotation with the TopGO R 800 package [120] with a Fisher exact test and an elim algorithm. 801 Synteny detection and multi-genome alignment of notothenioids 802 The same set of notothenioid genomes included in divergence time estimation was used to infer syntenic 803 relationships based on gene order conservation using GENESPACE v1.2.3 [121] with default 804 parameters and Progressive Cactus v2.9.7 [122] to perform a multi whole-genome alignment. Prior to 805 running Progressive Cactus, the genome assemblies were soft masked with RepeatMasker, using a 806 repeat library generated for each genome using RepeatModeler2 with default parameters. To guide the 807 whole-genome alignment we used the notothenioid -only subtree obtained with the constrained tree 808 search and the 1,000 randomly selected BUSCO genes. 809 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 28 We inferred insertion and deletion events along all branches of the species tree with the 810 halBranchMutations function of the HAL package [123]. Net genome gain rates per MY were calculated 811 by subtracting the total number of base pairs involved in deletion events from those involved in insertion 812 events and dividing the result by the branch length in time. Insertions inferred along the H. antarcticus 813 ancestors were then lifted off to the H. antarcticus Hap1 assembly using halLiftover (--noDupes). Due 814 to the fragmentation of the lifted insertions, we merged intervals closer than 100 bp when they derive 815 from the same ancestral insertion events. To detect putative TE -derived insertion events, lifted 816 annotations were intersected with the MCHelper-derived TE annotation requiring a reciprocal overlap 817 of the TE annotation and lifted insertion of 75%. PhyloFit from the PHAST package v1.5 [124] was 818 used under a REV substitution model to train an initial non-conserved model of evolution on the same 819 set of 4 -fold degenerate sites extracted from BUSCO genes that was used to estimate the neutral 820 substitution rate. PhastCons was then applied to a MAF representation of the whole -genome multiple 821 alignment, generated with cactus-hal2maf using H. antarcticus as the reference genome (--noAncestors 822 --refGenome HarAnt1.hap1 --dupeMode single), to refine the non -conserved model and train a 823 conserved model for each chromosome. Parameters from the per -chromosome conserved and non -824 conserved models were then averaged using PhyloBoot, and PhastCons was re -run to predict discrete 825 Conserved Elements (CEs). CEs closer than 5 bp were then merged into single elements. To identify 826 CEs shared by all analysed notothenioids, and those that emerged along the cryonotothenioid stem 827 branch, we used halAlignmentDepth to calculate per -base alignment depth for the H. antarcticus 828 genome. Elements covered over at least 75% of their length by all other notothenioids were considered 829 ancestrally present, whereas elements covered over at least 75% of their length only by 830 cryonotothenioids were considered to have been acquired along their stem branch. We assigned each 831 CE to different genomic features (exons, introns, promoters, transcription terminating site , and 832 intergenic regions) with the HOMER annotatePeaks.pl script [125]. GO term enrichment analyses was 833 performed similarly to what was performed for selection analyses but using all H. antarcticus genes as 834 background. 835 Supplementary Material 836 Supplementary Figures 837 Supplementary Tables 838 Data availability 839 The datasets generated in the present study have been submitted on NCBI and will become public 840 upon acceptance for publication. Biosample: SAMEA104132831. 841 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 29 Acknowledgments 842 We thank the Wellcome Sanger Institute DNA pipelines for support with sequencing data generation 843 of the H. antarcticus genome. We would like to thank the Rothera Marine Team for sample collection. 844 IB and JM were supported by the Centre for Translational Biodiversity Genomics (LOEWE -TBG) 845 funded through the program LOEWE –Landes-Offensive zur Entwicklung Wissenschaftlich -846 ökonomischer Exzellenz of Hesse's Ministry of Higher Education, Research, and the Arts. NF and DB 847 were supported by NIH grant R35 GM144336. 848 Competing Interests 849 The authors declare that they have no competing interests. 850 Author Contributions 851 I.B. conceptualised the study and supervised the work. I.B., J.M. designed analysis. A.D., M.S.C., N.F., 852 I.B., generated data. I.B., R.D., D.L.B., N.F. contributed resources, K.K. performed genome assembly, 853 T.M., J.M.D.W. performed genome curation, J.M. performed data analysis, with J.M.D.W input. J.M. 854 and I.B. wrote the original manuscript with contributions from M.S.C. All co -authors read, provided 855 feedback, and approved the manuscript. 856 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 30

857 1. Lecointre G, Améziane N, Boisselier M-C, Bonillo C, Busson F, Causse R, et al. Is the 858 Species Flock Concept Operational? The Antarctic Shelf Case. PLOS ONE [Internet]. 859 Public Library of Science; 2013 [cited 2026 Jan 9];8:e68787. 860 https://doi.org/10.1371/journal.pone.0068787 861 2. Eastman JT, Eakin RR. Checklist of the species of notothenioid fishes. Antarctic 862 Science [Internet]. 2021 [cited 2025 Nov 28];33:273–80. 863 https://doi.org/10.1017/S0954102020000632 864 3. Di Prisco G, Pisano E, Clarke A. Fishes of Antarctica: a biological overview Edited by 865 G. di Prisco, E. Pisano & A. Clarke Springer-Verlag, Milano (1998). 363 pages. Price DM 866 198. ISBN 88 470 0028 9. Antarctic Science [Internet]. 1999 [cited 2026 Jan 15];11:379–867 80. https://doi.org/10.1017/S0954102099210462 868 4. Chen L, DeVries AL, Cheng C-HC. Evolution of antifreeze glycoprotein gene from a 869 trypsinogen gene in Antarctic notothenioid fish. Proceedings of the National Academy 870 of Sciences [Internet]. Proceedings of the National Academy of Sciences; 1997 [cited 871 2026 Jan 14];94:3811–6. https://doi.org/10.1073/pnas.94.8.3811 872 5. Berthelot C, Clarke J, Desvignes T, William Detrich H III, Flicek P, Peck LS, et al. 873 Adaptation of Proteins to the Cold in Antarctic Fish: A Role for Methionine? Genome 874 Biol Evol [Internet]. 2019 [cited 2025 Nov 19];11:220–31. 875 https://doi.org/10.1093/gbe/evy262 876 6. Bista I, Wood JM, Desvignes T, McCarthy SA, Matschiner M, Ning Z, et al. Genomics of 877 cold adaptations in the Antarctic notothenioid fish radiation. Nature communications 878 [Internet]. Nature Publishing Group UK London; 2023 [cited 2025 Aug 25];14:3412. 879 https://www.nature.com/articles/s41467-023-38567-6. Accessed 25 Aug 2025 880 7. Cheng C-HC, Chen L. Evolution of an antifreeze glycoprotein. Nature [Internet]. 881 Nature Publishing Group; 1999 [cited 2026 Jan 15];401:443–4. 882 https://doi.org/10.1038/46721 883 8. Nicodemus-Johnson J, Silic S, Ghigliotti L, Pisano E, Cheng C-HC. Assembly of the 884 antifreeze glycoprotein/trypsinogen-like protease genomic locus in the Antarctic 885 toothfish Dissostichus mawsoni (Norman). Genomics [Internet]. 2011 [cited 2025 Nov 886 30];98:194–201. https://doi.org/10.1016/j.ygeno.2011.06.002 887 9. Rivera-Colón AG, Rayamajhi N, Minhas BF, Madrigal G, Bilyk KT, Yoon V, et al. 888 Genomics of Secondarily Temperate Adaptation in the Only Non-Antarctic Icefish. Mol 889 Biol Evol [Internet]. 2023 [cited 2026 Jan 14];40:msad029. 890 https://doi.org/10.1093/molbev/msad029 891 10. Ryder D, Stone D, Minardi D, Riley A, Avant J, Cross L, et al. De novo assembly and 892 annotation of the Patagonian toothfish (Dissostichus eleginoides) genome. BMC 893 Genomics [Internet]. 2024 [cited 2026 Jan 15];25:233. https://doi.org/10.1186/s12864-894 024-10141-4 895 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 31 11. Fraser KPP, Peck LS, Clark MS, Clarke A, Hill SL. Life in the freezer: protein 896 metabolism in Antarctic fish. R Soc Open Sci [Internet]. 2022 [cited 2026 Jan 897 15];9:211272. https://doi.org/10.1098/rsos.211272 898 12. Fraser KPP, Peck LS, Clark MS, Clarke A. A comparative study of tissue protein 899 synthesis rates in an Antarctic, Harpagifer antarcticus and a temperate, Lipophrys 900 pholis teleost. Comparative Biochemistry and Physiology Part A: Molecular & 901 Integrative Physiology [Internet]. 2024 [cited 2026 Jan 9];295:111650. 902 https://doi.org/10.1016/j.cbpa.2024.111650 903 13. Marty A-PM, Ward EN, Lamb JR, van Tartwijk FW, Peck LS, Clark MS, et al. A High-904 Resolution Microscopy System for Biological Studies of Cold-Adapted Species Under 905 Physiological Conditions. Small Methods [Internet]. 2025 [cited 2026 Mar 2];9:2401682. 906 https://doi.org/10.1002/smtd.202401682 907 14. Detrich HW, Stuart A, Schoenborn M, Parker SK, Methé BA, Amemiya CT. Genome 908 enablement of the notothenioidei: genome size estimates from 11 species and BAC 909 libraries from 2 Representative Taxa. Journal of Experimental Zoology Part B: Molecular 910 and Developmental Evolution [Internet]. 2010 [cited 2025 Nov 12];314B:369–81. 911 https://doi.org/10.1002/jez.b.21341 912 15. Chen L, Lu Y, Li W, Ren Y, Yu M, Jiang S, et al. The genomic basis for colonizing the 913 freezing Southern Ocean revealed by Antarctic toothfish and Patagonian robalo 914 genomes. Gigascience [Internet]. 2019 [cited 2026 Jan 14];8:giz016. 915 https://doi.org/10.1093/gigascience/giz016 916 16. Auvinet J, Graça P, Ghigliotti L, Pisano E, Dettaï A, Ozouf-Costaz C, et al. Insertion 917 Hot Spots of DIRS1 Retrotransposon and Chromosomal Diversifications among the 918 Antarctic Teleosts Nototheniidae. International Journal of Molecular Sciences 919 [Internet]. Multidisciplinary Digital Publishing Institute; 2019 [cited 2026 Jan 14];20:701. 920 https://doi.org/10.3390/ijms20030701 921 17. Coronado-Zamora M, González J. Transposable Elements and Adaptation. 922 Transposable Elements and Genome Evolution [Internet]. John Wiley & Sons, Ltd; 2024 923 [cited 2025 Nov 14]. p. 115–33. https://doi.org/10.1002/9781394312467.ch5 924 18. Bourque G, Burns KH, Gehring M, Gorbunova V, Seluanov A, Hammell M, et al. Ten 925 things you should know about transposable elements. Genome Biol [Internet]. 2018 926 [cited 2026 Feb 5];19:199. https://doi.org/10.1186/s13059-018-1577-z 927 19. Rey O, Danchin E, Mirouze M, Loot C, Blanchet S. Adaptation to Global Change: A 928 Transposable Element–Epigenetics Perspective. Trends in Ecology & Evolution 929 [Internet]. 2016 [cited 2026 Jan 14];31:514–26. 930 https://doi.org/10.1016/j.tree.2016.03.013 931 20. Xu H, Fang C, Xu W, Wang C, Song Y, Zhu C, et al. Evolution and genetic adaptation 932 of fishes to the deep sea. Cell [Internet]. Elsevier; 2025 [cited 2026 Jan 14];188:1393-933 1408.e13. https://doi.org/10.1016/j.cell.2025.01.002 934 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 32 21. Jarvis ED, Mirarab S, Aberer AJ, Li B, Houde P, Li C, et al. Whole-genome analyses 935 resolve early branches in the tree of life of modern birds. Science [Internet]. American 936 Association for the Advancement of Science; 2014 [cited 2026 Jan 14];346:1320–31. 937 https://doi.org/10.1126/science.1253451 938 22. Cicconardi F, Milanetti E, Pinheiro de Castro EC, Mazo-Vargas A, Van Belleghem 939 SM, Ruggieri AA, et al. Evolutionary dynamics of genome size and content during the 940 adaptive radiation of Heliconiini butterflies. Nat Commun [Internet]. Nature Publishing 941 Group; 2023 [cited 2026 Jan 14];14:5620. https://doi.org/10.1038/s41467-023-41412-5 942 23. Bista I, McCarthy SA, Wood J, Ning Z, Detrich Iii HW, Desvignes T, et al. The genome 943 sequence of the channel bull blenny, Cottoperca gobio (Günther, 1861). Wellcome 944 open research [Internet]. 2020 [cited 2025 Aug 25];5:148. 945 https://pmc.ncbi.nlm.nih.gov/articles/PMC7649722/. Accessed 25 Aug 2025 946 24. Pisano E, Ozouf-Costaz C. Chromosome change and the evolution in the Antarctic 947 fish suborder Notothenioidei. Antarctic Science [Internet]. 2000 [cited 2026 Jan 948 14];12:334–42. https://doi.org/10.1017/S0954102000000390 949 25. Kapitonov VV, Tempel S, Jurka J. Simple and fast classification of non-LTR 950 retrotransposons based on phylogeny of their RT domain protein sequences. Gene 951 [Internet]. 2009 [cited 2026 Jan 14];448:207–13. 952 https://doi.org/10.1016/j.gene.2009.07.019 953 26. Pisano E, Ghigliotti L. Ribosomal genes in notothenioid fishes: Focus on the 954 chromosomal organisation. Mar Genomics. 2009;2:75–80. 955 https://doi.org/10.1016/j.margen.2009.03.006 956 27. Pisano E, Ozouf-Costaz C, Prirodina V. Chromosome Diversification in Antarctic 957 Fish (Notothenioidei). In: Di Prisco G, Pisano E, Clarke A, editors. Fishes of Antarctica: A 958 biological overview [Internet]. Milano: Springer Milan; 1998 [cited 2026 Jan 9]. p. 275–959 85. https://doi.org/10.1007/978-88-470-2157-0_24 960 28. Miller KG, Browning JV, Schmelz WJ, Kopp RE, Mountain GS, Wright JD. Cenozoic 961 sea-level and cryospheric evolution from deep-sea geochemical and continental 962 margin records. Science Advances [Internet]. American Association for the 963 Advancement of Science; 2020 [cited 2026 Jan 14];6:eaaz1346. 964 https://doi.org/10.1126/sciadv.aaz1346 965 29. Stankovic A, Spalik K, Kamler E, Borsuk P, Weglenski P. Recent origin of sub-966 Antarctic notothenioids. Polar Biol [Internet]. 2002 [cited 2026 Jan 14];25:203–5. 967 https://doi.org/10.1007/s00300-001-0327-x 968 30. Collins M, Brickle P, Brown J, Belchier M. The Patagonian Toothfish: Biology, Ecology 969 and Fishery. Advances in Marine Biology [Internet]. Academic Press; 2010 [cited 2026 970 Jan 14]. p. 227–300. https://doi.org/10.1016/B978-0-12-381015-1.00004-6 971 31. Near TJ, Dornburg A, Harrington RC, Oliveira C, Pietsch TW, Thacker CE, et al. 972 Identification of the notothenioid sister lineage illuminates the biogeographic history of 973 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 33 an Antarctic adaptive radiation. BMC Evol Biol [Internet]. 2015 [cited 2026 Jan 974 14];15:109. https://doi.org/10.1186/s12862-015-0362-9 975 32. Detrich HW, Parker SK, Williams RC, Nogales E, Downing KH. Cold Adaptation of 976 Microtubule Assembly and Dynamics: STRUCTURAL INTERPRETATION OF PRIMARY 977 SEQUENCE CHANGES PRESENT IN THE α- AND β-TUBULINS OF ANTARCTIC FISHES *. 978 Journal of Biological Chemistry [Internet]. Elsevier; 2000 [cited 2026 Jan 14];275:37038–979 47. https://doi.org/10.1074/jbc.M005699200 980 33. Kowalczyk A, Chikina M, Clark NL. A cautionary tale on proper use of branch-site 981 models to detect convergent positive selection [Internet]. bioRxiv; 2021 [cited 2026 Feb 982 5]. p. 2021.10.26.465984. https://doi.org/10.1101/2021.10.26.465984 983 34. Rayamajhi N, Rivera-Colón AG, Minhas BF, Cheng CHC, Catchen JM. The genome of 984 the cryopelagic Antarctic bald notothen, Trematomus borchgrevinki. G3 985 Genes|Genomes|Genetics [Internet]. 2025 [cited 2026 Jan 14];15:jkae267. 986 https://doi.org/10.1093/g3journal/jkae267 987 35. Liu S, Shi C, Chen C, Tan Y, Tian Y, Macqueen DJ, et al. Haplotype-resolved 988 genomes provide insights into the origins and functional significance of genome 989 diversity in bivalves. Cell Reports [Internet]. 2025 [cited 2026 Jan 14];44:115697. 990 https://doi.org/10.1016/j.celrep.2025.115697 991 36. Rosani U, Gerdol M, Krupovic M. The highly dynamic pangenome of basal chordates 992 is enriched in defence and immunity genes and is inherited following the Mendelian 993 law. PLOS Genetics [Internet]. Public Library of Science; 2025 [cited 2026 Feb 994 5];21:e1011833. https://doi.org/10.1371/journal.pgen.1011833 995 37. Audano PA, Sulovari A, Graves-Lindsay TA, Cantsilieris S, Sorensen M, Welch AE, et 996 al. Characterizing the Major Structural Variant Alleles of the Human Genome. Cell 997 [Internet]. Elsevier; 2019 [cited 2026 Jan 14];176:663-675.e19. 998 https://doi.org/10.1016/j.cell.2018.12.019 999 38. Chang N-C, Rovira Q, Wells J, Feschotte C, Vaquerizas JM. Zebrafish transposable 1000 elements show extensive diversification in age, genomic distribution, and 1001 developmental expression. Genome Res [Internet]. Cold Spring Harbor Lab; 2022 [cited 1002 2026 Jan 14];32:1408–23. https://doi.org/10.1101/gr.275655.121 1003 39. Li H, Bloom JM, Farjoun Y, Fleharty M, Gauthier L, Neale B, et al. A synthetic-diploid 1004 benchmark for accurate variant-calling evaluation. Nat Methods [Internet]. Nature 1005 Publishing Group; 2018 [cited 2026 Jan 14];15:595–7. https://doi.org/10.1038/s41592-1006 018-0054-7 1007 40. Gao Y, Yang X, Chen H, Tan X, Yang Z, Deng L, et al. A pangenome reference of 36 1008 Chinese populations. Nature [Internet]. Nature Publishing Group; 2023 [cited 2026 Jan 1009 14];619:112–21. https://doi.org/10.1038/s41586-023-06173-7 1010 41. Li Q, Qiao X, Li L, Gu C, Yin H, Qi K, et al. Haplotype-resolved T2T genome 1011 assemblies and pangenome graph of pear reveal diverse patterns of allele-specific 1012 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 34 expression and the genomic basis of fruit quality traits. Plant Communications 1013 [Internet]. 2024 [cited 2026 Jan 14];5:101000. 1014 https://doi.org/10.1016/j.xplc.2024.101000 1015 42. Sierra P, Durbin R. Identification of transposable element families from pangenome 1016 polymorphisms. Mobile DNA [Internet]. 2024 [cited 2026 Jan 14];15:13. 1017 https://doi.org/10.1186/s13100-024-00323-y 1018 43. Han S, Dias GB, Basting PJ, Viswanatha R, Perrimon N, Bergman CM. Local 1019 assembly of long reads enables phylogenomics of transposable elements in a polyploid 1020 cell line. Nucleic Acids Res [Internet]. 2022 [cited 2026 Jan 14];50:e124. 1021 https://doi.org/10.1093/nar/gkac794 1022 44. Boissinot S, Sookdeo A. The Evolution of LINE-1 in Vertebrates. Genome Biol Evol 1023 [Internet]. 2016 [cited 2026 Jan 14];8:3485–507. https://doi.org/10.1093/gbe/evw247 1024 45. Martelossi J, Nicolini F, Subacchi S, Pasquale D, Ghiselli F, Luchetti A. Multiple and 1025 diversified transposon lineages contribute to early and recent bivalve genome 1026 evolution. BMC Biol [Internet]. 2023 [cited 2026 Jan 14];21:145. 1027 https://doi.org/10.1186/s12915-023-01632-z 1028 46. Davis KE, Hill J, Astrop TI, Wills MA. Global cooling as a driver of diversification in a 1029 major marine clade. Nat Commun [Internet]. Nature Publishing Group; 2016 [cited 2026 1030 Jan 14];7:13003. https://doi.org/10.1038/ncomms13003 1031 47. Chénais B, Caruso A, Hiard S, Casse N. The impact of transposable elements on 1032 eukaryotic genomes: From genome size increase to genetic adaptation to stressful 1033 environments. Gene [Internet]. 2012 [cited 2026 Jan 14];509:7–15. 1034 https://doi.org/10.1016/j.gene.2012.07.042 1035 48. Chen S, Yu M, Chu X, Li W, Yin X, Chen L. Cold-induced retrotransposition of fish 1036 LINEs. Journal of Genetics and Genomics [Internet]. 2017 [cited 2026 Jan 14];44:385–1037 94. https://doi.org/10.1016/j.jgg.2017.07.002 1038 49. Pisano E, Ozouf-Costaz C, Prirodina V. Chromosome Diversification in Antarctic 1039 Fish (Notothenioidei). In: Di Prisco G, Pisano E, Clarke A, editors. Fishes of Antarctica: A 1040 biological overview [Internet]. Milano: Springer Milan; 1998 [cited 2026 Jan 15]. p. 275–1041 85. https://doi.org/10.1007/978-88-470-2157-0_24 1042 50. Amores A, Wilson CA, Allard CAH, Detrich HW III, Postlethwait JH. Cold Fusion: 1043 Massive Karyotype Evolution in the Antarctic Bullhead Notothen Notothenia coriiceps. 1044 G3 Genes|Genomes|Genetics [Internet]. 2017 [cited 2026 Mar 9];7:2195–207. 1045 https://doi.org/10.1534/g3.117.040063 1046 51. Gerton JL. A working model for the formation of Robertsonian chromosomes. J Cell 1047 Sci [Internet]. 2024 [cited 2026 Jan 11];137:jcs261912. 1048 https://doi.org/10.1242/jcs.261912 1049 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 35 52. Read LR, Raynard SJ, Rukść A, Baker MD. Gene repeat expansion and contraction by 1050 spontaneous intrachromosomal homologous recombination in mammalian cells. 1051 Nucleic Acids Res. 2004;32:1184–96. https://doi.org/10.1093/nar/gkh280 1052 53. Raymond JA, DeVries AL. Adsorption inhibition as a mechanism of freezing 1053 resistance in polar fishes. Proceedings of the National Academy of Sciences [Internet]. 1054 Proceedings of the National Academy of Sciences; 1977 [cited 2026 Jan 21];74:2589–1055 93. https://doi.org/10.1073/pnas.74.6.2589 1056 54. DeVries AL. The role of antifreeze glycopeptides and peptides in the freezing 1057 avoidance of antarctic fishes. Comparative Biochemistry and Physiology Part B: 1058 Comparative Biochemistry [Internet]. 1988 [cited 2026 Jan 21];90:611–21. 1059 https://doi.org/10.1016/0305-0491(88)90302-1 1060 55. Schrag JD, Devries AL. The effect of freezing rate on the cooperativity of antifreeze 1061 glycopeptides. Comparative Biochemistry and Physiology Part A: Physiology [Internet]. 1062 1983 [cited 2026 Jan 21];74:381–5. https://doi.org/10.1016/0300-9629(83)90619-9 1063 56. Frandsen PB, Hotaling S, Powell A, Heckenhauer J, Kawahara AY, Baker RH, et al. 1064 Allelic resolution of insect and spider silk genes reveals hidden genetic diversity. 1065 Proceedings of the National Academy of Sciences [Internet]. Proceedings of the 1066 National Academy of Sciences; 2023 [cited 2026 Jan 14];120:e2221528120. 1067 https://doi.org/10.1073/pnas.2221528120 1068 57. Arakawa K, Kono N, Malay AD, Tateishi A, Ifuku N, Masunaga H, et al. 1000 spider 1069 silkomes: Linking sequences to silk physical properties. Science Advances [Internet]. 1070 American Association for the Advancement of Science; 2022 [cited 2026 Jan 1071 14];8:eabo6043. https://doi.org/10.1126/sciadv.abo6043 1072 58. Miya T, Gon O, Mwale M, Cheng C-HC. Multiple independent reduction or loss of 1073 antifreeze trait in low Antarctic and sub-Antarctic notothenioid fishes. Antarctic 1074 Science [Internet]. 2016 [cited 2025 Nov 12];28:17–28. 1075 https://doi.org/10.1017/S0954102015000413 1076 59. Abele D, Puntarulo S. Formation of reactive species and induction of antioxidant 1077 defence systems in polar and temperate marine invertebrates and fish. Comparative 1078 Biochemistry and Physiology Part A: Molecular & Integrative Physiology [Internet]. 2004 1079 [cited 2026 Jan 14];138:405–15. https://doi.org/10.1016/j.cbpb.2004.05.013 1080 60. Peck LS. A Cold Limit to Adaptation in the Sea. Trends in Ecology & Evolution 1081 [Internet]. Elsevier; 2016 [cited 2026 Jan 11];31:13–26. 1082 https://doi.org/10.1016/j.tree.2015.09.014 1083 61. Carpenter JH. New Measurements of Oxygen Solubility in Pure and Natural Water. 1084 Limnology and Oceanography [Internet]. 1966 [cited 2026 Jan 11];11:264–77. 1085 https://doi.org/10.4319/lo.1966.11.2.0264 1086 62. Olmos Y, Sánchez-Gómez FJ, Wild B, García-Quintans N, Cabezudo S, Lamas S, et 1087 al. SirT1 Regulation of Antioxidant Genes Is Dependent on the Formation of a 1088 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 36 FoxO3a/PGC-1α Complex. Antioxidants & Redox Signaling [Internet]. Mary Ann Liebert, 1089 Inc., publishers; 2013 [cited 2026 Jan 14];19:1507–21. 1090 https://doi.org/10.1089/ars.2012.4713 1091 63. Ferko M, Alanova P, Janko D, Opletalova B, Andelova N. Mitochondrial 1092 Peroxiredoxins and Monoamine Oxidase-A: Dynamic Regulators of ROS Signaling in 1093 Cardioprotection. Physiol Res [Internet]. 2024 [cited 2026 Jan 14];73:887–900. 1094 https://doi.org/10.33549/physiolres.935513 1095 64. Bilyk KT, Zhuang X, Papetti C. Positive and Relaxed Selective Pressures Have Both 1096 Strongly Influenced the Evolution of Cryonotothenioid Fishes during Their Radiation in 1097 the Freezing Southern Ocean. Genome Biol Evol [Internet]. 2023 [cited 2026 Jan 1098 23];15:evad049. https://doi.org/10.1093/gbe/evad049 1099 65. Shin SC, Ahn DH, Kim SJ, Pyo CW, Lee H, Kim M-K, et al. The genome sequence of 1100 the Antarctic bullhead notothen reveals evolutionary adaptations to a cold 1101 environment. Genome Biol [Internet]. 2014 [cited 2026 Jan 14];15:468. 1102 https://doi.org/10.1186/s13059-014-0468-1 1103 66. Chen Z, Cheng C-HC, Zhang J, Cao L, Chen L, Zhou L, et al. Transcriptomic and 1104 genomic evolution under constant cold in Antarctic notothenioid fish. Proceedings of 1105 the National Academy of Sciences [Internet]. Proceedings of the National Academy of 1106 Sciences; 2008 [cited 2025 Nov 19];105:12944–9. 1107 https://doi.org/10.1073/pnas.0802432105 1108 67. Clark MS, Thorne MAS, Toullec J-Y, Meng Y, Guan LL, Peck LS, et al. Antarctic Krill 1109 454 Pyrosequencing Reveals Chaperone and Stress Transcriptome. PLOS ONE 1110 [Internet]. Public Library of Science; 2011 [cited 2026 Jan 11];6:e15919. 1111 https://doi.org/10.1371/journal.pone.0015919 1112 68. Abele D, Tesch C, Wencke P, Pörtner HO. How does oxidative stress relate to 1113 thermal tolerance in the Antarctic bivalve Yoldia eightsi? Antarctic Science [Internet]. 1114 2001 [cited 2026 Jan 11];13:111–8. https://doi.org/10.1017/S0954102001000189 1115 69. Fraser KPP, Rogers AD. Protein Metabolism in Marine Animals: The Underlying 1116 Mechanism of Growth. Advances in Marine Biology [Internet]. Academic Press; 2007 1117 [cited 2026 Jan 11]. p. 267–362. https://doi.org/10.1016/S0065-2881(06)52003-6 1118 70. Todgham AE, Hoaglund EA, Hofmann GE. Is cold the new hot? Elevated ubiquitin-1119 conjugated protein levels in tissues of Antarctic fish as evidence for cold-denaturation 1120 of proteins in vivo. J Comp Physiol B [Internet]. 2007 [cited 2026 Jan 11];177:857–66. 1121 https://doi.org/10.1007/s00360-007-0183-2 1122 71. Hofmann GE, Lund SG, Place SP, Whitmer AC. Some like it hot, some like it cold: 1123 the heat shock response is found in New Zealand but not Antarctic notothenioid fishes. 1124 Journal of Experimental Marine Biology and Ecology [Internet]. 2005 [cited 2026 Jan 1125 11];316:79–89. https://doi.org/10.1016/j.jembe.2004.10.007 1126 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 37 72. Place SP, Zippay ML, Hofmann GE. Constitutive roles for inducible genes: evidence 1127 for the alteration in expression of the inducible hsp70 gene in Antarctic notothenioid 1128 fishes. American Journal of Physiology-Regulatory, Integrative and Comparative 1129 Physiology [Internet]. American Physiological Society; 2004 [cited 2026 Jan 11]; 1130 https://doi.org/10.1152/ajpregu.00223.2004 1131 73. Todgham AE, Crombie TA, Hofmann GE. The effect of temperature adaptation on 1132 the ubiquitin–proteasome pathway in notothenioid fishes. J Exp Biol [Internet]. 2017 1133 [cited 2026 Jan 11];220:369–78. https://doi.org/10.1242/jeb.145946 1134 74. Strickland M, Moll R, Cornwell C, Smith M, Howard C. Sanger Tree of Life HMW DNA 1135 Extraction: Manual MagAttract. 2023 [cited 2026 Feb 16]; 1136 https://www.protocols.io/view/sanger-tree-of-life-hmw-dna-extraction-manual-maga-1137 cy3ixyke. Accessed 16 Feb 2026 1138 75. Cheng H, Concepcion GT, Feng X, Zhang H, Li H. Haplotype-resolved de novo 1139 assembly using phased assembly graphs with hifiasm. Nat Methods [Internet]. Nature 1140 Publishing Group; 2021 [cited 2026 Jan 15];18:170–5. https://doi.org/10.1038/s41592-1141 020-01056-5 1142 76. Vasimuddin Md, Misra S, Li H, Aluru S. Efficient Architecture-Aware Acceleration of 1143 BWA-MEM for Multicore Systems. 2019 IEEE International Parallel and Distributed 1144 Processing Symposium (IPDPS) [Internet]. 2019 [cited 2026 Jan 15]. p. 314–24. 1145 https://doi.org/10.1109/IPDPS.2019.00041 1146 77. Zhou C, McCarthy SA, Durbin R. YaHS: yet another Hi-C scaffolding tool. 1147 Bioinformatics [Internet]. 2023 [cited 2026 Jan 14];39:btac808. 1148 https://doi.org/10.1093/bioinformatics/btac808 1149 78. Flynn JM, Hubley R, Goubert C, Rosen J, Clark AG, Feschotte C, et al. 1150 RepeatModeler2 for automated genomic discovery of transposable element families. 1151 Proc Natl Acad Sci USA [Internet]. 2020 [cited 2025 Dec 9];117:9451–7. 1152 https://doi.org/10.1073/pnas.1921046117 1153 79. Lin Y, Ye C, Li X, Chen Q, Wu Y, Zhang F, et al. quarTeT: a telomere-to-telomere 1154 toolkit for gap-free genome assembly and centromeric repeat identification. Hortic Res 1155 [Internet]. 2023 [cited 2026 Feb 5];10:uhad127. https://doi.org/10.1093/hr/uhad127 1156 80. Seemann T. tseemann/barrnap [Internet]. 2026 [cited 2026 Mar 5]. 1157 https://github.com/tseemann/barrnap. Accessed 5 Mar 2026 1158 81. Brown MR, Manuel Gonzalez de La Rosa P, Blaxter M. tidk: a toolkit to rapidly 1159 identify telomeric repeats from genomic datasets. Bioinformatics [Internet]. 2025 [cited 1160 2026 Jan 14];41:btaf049. https://doi.org/10.1093/bioinformatics/btaf049 1161 82. Gabriel L, Brůna T, Hoff KJ, Ebel M, Lomsadze A, Borodovsky M, et al. BRAKER3: 1162 Fully automated genome annotation using RNA-seq and protein evidence with 1163 GeneMark-ETP, AUGUSTUS, and TSEBRA. Genome Res [Internet]. Cold Spring Harbor 1164 Lab; 2024 [cited 2026 Jan 14];34:769–77. https://doi.org/10.1101/gr.278090.123 1165 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 38 83. Hoff KJ, Stanke M. Predicting Genes in Single Genomes with AUGUSTUS. Current 1166 Protocols in Bioinformatics [Internet]. 2019 [cited 2026 Jan 14];65:e57. 1167 https://doi.org/10.1002/cpbi.57 1168 84. Brůna T, Lomsadze A, Borodovsky M. GeneMark-EP+: eukaryotic gene prediction 1169 with self-training in the space of genes and proteins. NAR Genom Bioinform [Internet]. 1170 2020 [cited 2026 Jan 14];2:lqaa026. https://doi.org/10.1093/nargab/lqaa026 1171 85. Kim D, Paggi JM, Park C, Bennett C, Salzberg SL. Graph-based genome alignment 1172 and genotyping with HISAT2 and HISAT-genotype. Nat Biotechnol [Internet]. Nature 1173 Publishing Group; 2019 [cited 2026 Jan 14];37:907–15. https://doi.org/10.1038/s41587-1174 019-0201-4 1175 86. Manni M, Berkeley MR, Seppey M, Simão FA, Zdobnov EM. BUSCO Update: Novel 1176 and Streamlined Workflows along with Broader and Deeper Phylogenetic Coverage for 1177 Scoring of Eukaryotic, Prokaryotic, and Viral Genomes. Mol Biol Evol [Internet]. 2021 1178 [cited 2026 Jan 14];38:4647–54. https://doi.org/10.1093/molbev/msab199 1179 87. Shumate A, Salzberg SL. Liftoff: accurate mapping of gene annotations. 1180 Bioinformatics [Internet]. 2021 [cited 2026 Jan 14];37:1639–43. 1181 https://doi.org/10.1093/bioinformatics/btaa1016 1182 88. Cantalapiedra CP, Hernández-Plaza A, Letunic I, Bork P, Huerta-Cepas J. eggNOG-1183 mapper v2: Functional Annotation, Orthology Assignments, and Domain Prediction at 1184 the Metagenomic Scale. Mol Biol Evol [Internet]. 2021 [cited 2026 Jan 14];38:5825–9. 1185 https://doi.org/10.1093/molbev/msab293 1186 89. Keilwagen J, Hartung F, Grau J. GeMoMa: Homology-Based Gene Prediction Utilizing 1187 Intron Position Conservation and RNA-seq Data. In: Kollmar M, editor. Gene Prediction: 1188

and Protocols [Internet]. New York, NY: Springer; 2019 [cited 2026 Jan 14]. p. 1189 161–77. https://doi.org/10.1007/978-1-4939-9173-0_9 1190 90. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search 1191 tool. Journal of Molecular Biology [Internet]. 1990 [cited 2026 Jan 14];215:403–10. 1192 https://doi.org/10.1016/S0022-2836(05)80360-2 1193 91. Madrigal G, Minhas BF, Catchen J. Klumpy: A tool to evaluate the integrity of long-1194 read genome assemblies and illusive sequence motifs. Molecular Ecology Resources 1195 [Internet]. 2025 [cited 2026 Jan 14];25:e13982. https://doi.org/10.1111/1755-1196 0998.13982 1197 92. Katoh K, Standley DM. MAFFT Multiple Sequence Alignment Software Version 7: 1198 Improvements in Performance and Usability. Mol Biol Evol [Internet]. 2013 [cited 2026 1199 Jan 14];30:772–80. https://doi.org/10.1093/molbev/mst010 1200 93. Minh BQ, Schmidt HA, Chernomor O, Schrempf D, Woodhams MD, von Haeseler A, 1201 et al. IQ-TREE 2: New Models and Efficient Methods for Phylogenetic Inference in the 1202 Genomic Era. Mol Biol Evol [Internet]. 2020 [cited 2026 Jan 14];37:1530–4. 1203 https://doi.org/10.1093/molbev/msaa015 1204 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 39 94. Hoang DT, Chernomor O, von Haeseler A, Minh BQ, Vinh LS. UFBoot2: Improving the 1205 Ultrafast Bootstrap Approximation. Mol Biol Evol [Internet]. 2018 [cited 2026 Jan 1206 14];35:518–22. https://doi.org/10.1093/molbev/msx281 1207 95. Kalyaanamoorthy S, Minh BQ, Wong TKF, von Haeseler A, Jermiin LS. ModelFinder: 1208 fast model selection for accurate phylogenetic estimates. Nat Methods [Internet]. 1209 Nature Publishing Group; 2017 [cited 2026 Jan 14];14:587–9. 1210 https://doi.org/10.1038/nmeth.4285 1211 96. Mistry J, Chuguransky S, Williams L, Qureshi M, Salazar GA, Sonnhammer ELL, et al. 1212 Pfam: The protein families database in 2021. Nucleic Acids Res [Internet]. 2021 [cited 1213 2026 Jan 15];49:D412–9. https://doi.org/10.1093/nar/gkaa913 1214 97. Fu L, Niu B, Zhu Z, Wu S, Li W. CD-HIT: accelerated for clustering the next-1215 generation sequencing data. Bioinformatics [Internet]. 2012 [cited 2026 Jan 1216 15];28:3150–2. https://doi.org/10.1093/bioinformatics/bts565 1217 98. Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T. trimAl: a tool for automated 1218 alignment trimming in large-scale phylogenetic analyses. Bioinformatics [Internet]. 1219 2009 [cited 2026 Jan 14];25:1972–3. https://doi.org/10.1093/bioinformatics/btp348 1220 99. Platt RN II, Blanco-Berdugo L, Ray DA. Accurate Transposable Element Annotation 1221 Is Vital When Analyzing New Genome Assemblies. Genome Biol Evol [Internet]. 2016 1222 [cited 2025 Dec 3];8:403–10. https://doi.org/10.1093/gbe/evw009 1223 100. Goubert C, Craig RJ, Bilat AF, Peona V, Vogan AA, Protasio AV. A beginner’s guide 1224 to manual curation of transposable elements. Mobile DNA [Internet]. 2022 [cited 2025 1225 Dec 3];13:7. https://doi.org/10.1186/s13100-021-00259-7 1226 101. Peona V, Martelossi J, Almojil D, Bocharkina J, Brännström I, Brown M, et al. 1227 Teaching transposon classification as a means to crowd source the curation of repeat 1228 annotation – a tardigrade perspective. Mobile DNA [Internet]. 2024 [cited 2026 Jan 1229 15];15:10. https://doi.org/10.1186/s13100-024-00319-8 1230 102. Li H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 1231 [Internet]. 2018 [cited 2025 Oct 27];34:3094–100. 1232 https://doi.org/10.1093/bioinformatics/bty191 1233 103. Goel M, Sun H, Jiao W-B, Schneeberger K. SyRI: finding genomic rearrangements 1234 and local sequence differences from whole-genome assemblies. Genome Biol 1235 [Internet]. 2019 [cited 2026 Jan 14];20:277. https://doi.org/10.1186/s13059-019-1911-0 1236 104. Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic 1237 features. Bioinformatics [Internet]. 2010 [cited 2026 Jan 14];26:841–2. 1238 https://doi.org/10.1093/bioinformatics/btq033 1239 105. Bates D, Mächler M, Bolker B, Walker S. Fitting Linear Mixed-Effects Models Using 1240 lme4. Journal of Statistical Software [Internet]. 2015 [cited 2026 Jan 14];67:1–48. 1241 https://doi.org/10.18637/jss.v067.i01 1242 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 40 106. Smolka M, Paulin LF, Grochowski CM, Horner DW, Mahmoud M, Behera S, et al. 1243 Detection of mosaic and population-level structural variants with Sniffles2. Nat 1244 Biotechnol [Internet]. Nature Publishing Group; 2024 [cited 2026 Jan 14];42:1571–80. 1245 https://doi.org/10.1038/s41587-023-02024-y 1246 107. Jeffares DC, Jolly C, Hoti M, Speed D, Shaw L, Rallis C, et al. Transient structural 1247 variations have strong effects on quantitative traits and reproductive isolation in fission 1248 yeast. Nat Commun [Internet]. Nature Publishing Group; 2017 [cited 2026 Jan 1249 14];8:14061. https://doi.org/10.1038/ncomms14061 1250 108. Bista I, Collins M. The genome sequence of the marbled rockcod, Notothenia rossii 1251 Richardson, 1844. Wellcome Open Res [Internet]. 2024 [cited 2025 Nov 15];9:227. 1252 https://doi.org/10.12688/wellcomeopenres.21270.1 1253 109. Cheng C-HC, Rivera-Colón AG, Minhas BF, Wilson L, Rayamajhi N, Vargas-Chacoff 1254 L, et al. Chromosome-Level Genome Assembly and Circadian Gene Repertoire of the 1255 Patagonia Blennie Eleginops maclovinus—The Closest Ancestral Proxy of Antarctic 1256 Cryonotothenioids. Genes [Internet]. Multidisciplinary Digital Publishing Institute; 2023 1257 [cited 2026 Jan 15];14:1196. https://doi.org/10.3390/genes14061196 1258 110. Jae Lee S, Kim J-H, Jo E, Choi E, Kim J, Choi S-G, et al. Chromosomal assembly of 1259 the Antarctic toothfish (Dissostichus mawsoni) genome using third-generation DNA 1260 sequencing and Hi-C technology. Zool Res [Internet]. 2021 [cited 2026 Jan 15];42:124–1261 9. https://doi.org/10.24272/j.issn.2095-8137.2020.264 1262 111. Lee SJ, Kim J, Choi EK, Jo E, Cho M, Kim J-H, et al. A chromosome-level reference 1263 genome of the Antarctic blackfin icefish Chaenocephalus aceratus. Sci Data [Internet]. 1264 Nature Publishing Group; 2023 [cited 2026 Jan 15];10:657. 1265 https://doi.org/10.1038/s41597-023-02561-w 1266 112. Rayamajhi N, Rivera-Colón AG, Minhas BF, Cheng CHC, Catchen JM. The genome 1267 of the cryopelagic Antarctic bald notothen, Trematomus borchgrevinki. G3 1268 Genes|Genomes|Genetics [Internet]. 2025 [cited 2026 Jan 15];15:jkae267. 1269 https://doi.org/10.1093/g3journal/jkae267 1270 113. Jo E, Choi S, Lee SJ, Kim J, Choi EK, Cho M, et al. Chromosome-level genome 1271 assembly and annotation of the Antarctica whitefin plunderfish Pogonophryne 1272 albipinna. Sci Data [Internet]. Nature Publishing Group; 2023 [cited 2026 Jan 1273 15];10:891. https://doi.org/10.1038/s41597-023-02811-x 1274 114. To T-H, Jung M, Lycett S, Gascuel O. Fast Dating Using Least-Squares Criteria and 1275 Algorithms. Syst Biol [Internet]. 2016 [cited 2026 Jan 15];65:82–97. 1276 https://doi.org/10.1093/sysbio/syv068 1277 115. Emms DM, Kelly S. OrthoFinder: phylogenetic orthology inference for comparative 1278 genomics. Genome Biol [Internet]. 2019 [cited 2025 Dec 1];20:238. 1279 https://doi.org/10.1186/s13059-019-1832-y 1280 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint 41 116. Willson J, Roddur MS, Liu B, Zaharias P, Warnow T. DISCO: Species Tree Inference 1281 using Multicopy Gene Family Tree Decomposition. Syst Biol [Internet]. 2022 [cited 2026 1282 Jan 14];71:610–29. https://doi.org/10.1093/sysbio/syab070 1283 117. Löytynoja A. Phylogeny-aware alignment with PRANK. In: Russell DJ, editor. 1284 Multiple Sequence Alignment Methods [Internet]. Totowa, NJ: Humana Press; 2014 1285 [cited 2026 Jan 15]. p. 155–70. https://doi.org/10.1007/978-1-62703-646-7_10 1286 118. Talavera G, Castresana J. Improvement of Phylogenies after Removing Divergent 1287 and Ambiguously Aligned Blocks from Protein Sequence Alignments. Syst Biol 1288 [Internet]. 2007 [cited 2026 Jan 15];56:564–77. 1289 https://doi.org/10.1080/10635150701472164 1290 119. Smith MD, Wertheim JO, Weaver S, Murrell B, Scheffler K, Kosakovsky Pond SL. 1291 Less Is More: An Adaptive Branch-Site Random Effects Model for Efficient Detection of 1292 Episodic Diversifying Selection. Mol Biol Evol [Internet]. 2015 [cited 2026 Jan 1293 15];32:1342–53. https://doi.org/10.1093/molbev/msv022 1294 120. Alexa A, Rahnenführer J, Marini F. topGO [Internet]. Bioconductor. [cited 2026 Jan 1295 15]. http://bioconductor.org/packages/topGO/. Accessed 15 Jan 2026 1296 121. Lovell JT, Sreedasyam A, Schranz ME, Wilson M, Carlson JW, Harkess A, et al. 1297 GENESPACE tracks regions of interest and gene copy number variation across multiple 1298 genomes. Weigel D, editor. eLife [Internet]. eLife Sciences Publications, Ltd; 2022 [cited 1299 2026 Jan 15];11:e78526. https://doi.org/10.7554/eLife.78526 1300 122. Armstrong J, Hickey G, Diekhans M, Fiddes IT, Novak AM, Deran A, et al. 1301 Progressive Cactus is a multiple-genome aligner for the thousand-genome era. Nature 1302 [Internet]. Nature Publishing Group; 2020 [cited 2026 Jan 15];587:246–51. 1303 https://doi.org/10.1038/s41586-020-2871-y 1304 123. Hickey G, Paten B, Earl D, Zerbino D, Haussler D. HAL: a hierarchical format for 1305 storing and analyzing multiple genome alignments. Bioinformatics. 2013;29:1341–2. 1306 https://doi.org/10.1093/bioinformatics/btt128 1307 124. Hubisz MJ, Pollard KS, Siepel A. PHAST and RPHAST: phylogenetic analysis with 1308 space/time models. Brief Bioinform [Internet]. 2011 [cited 2026 Feb 5];12:41–51. 1309 https://doi.org/10.1093/bib/bbq072 1310 125. Heinz S, Benner C, Spann N, Bertolino E, Lin YC, Laslo P, et al. Simple 1311 Combinations of Lineage-Determining Transcription Factors Prime cis-Regulatory 1312 Elements Required for Macrophage and B Cell Identities. Molecular Cell [Internet]. 1313 Elsevier; 2010 [cited 2026 Feb 5];38:576–89. 1314 https://doi.org/10.1016/j.molcel.2010.05.004 1315 1316 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 23, 2026. ; https://doi.org/10.64898/2026.04.21.719633doi: bioRxiv preprint