Fitting together the evolutionary puzzle pieces of the Immunoglobulin T gene from Antarctic fishes

: Cryonotothenioidea is the main group of fishes that thrive in the extremely cold Antarctic environment, thanks to the acquisition of peculiar morphological, physiological and molecular adaptations. We have previously disclosed that IgM, the main immunoglobulin isotype in teleosts, display typical cold-adapted features. Recently, we have analyzed the gene encoding the heavy chain constant region (CH) of the IgT isotype from the Antarctic teleost Trematomus bernacchii (family Nototheniidae), characterized by the near-complete deletion of the CH2 domain. Here, we aimed to track the loss of the CH2 domain along notothenioid phylogeny and to identify its ancestral origins. To this end, we obtained the IgT gene sequences from several species belonging to the Antarctic families Nototheniidae, Bathydraconidae and Artedidraconidae. All species display a CH2 remnant of variable size, encoded by a short Cτ2 exon, which retains functional splicing sites and therefore is included in the mature transcript. We also considered representative species from the three non-Antarctic families: Eleginopsioidea ( Eleginops maclovinus ), Pseudaphritioidea ( Pseudaphritis urvillii ) and Bovichtidae ( Bovichtus diacanthus and Cottoperca gobio ). Even though only E. maclovinus , the sister taxa of Cryonotothenioidea, shared the partial loss of Cτ2 , the other non-Antarctic notothenioid species displayed early molecular signatures of this event. These results shed light on the evolutionary path that underlies the origins of this remarkable gene structural modification. compared with both non-Antarctic (40%) and temperate species (47%), suggesting a significant role of glycosylation for cold-adapted proteins.

The first report of an IgT CH region gene in a teleost species living under extreme conditions dates back to 2015, when our group reported the unprecedented case of a nearly complete truncation of the CH2 domain in the Antarctic fish Trematomus bernacchii [18]. Two out of the three variants identified in this species, termed Long (TbeL) and Short (TbeS), respectively) were encoded by alleles characterized by a large deletion and only displayed a short remnant of the Cτ2 exon, which was entirely skipped by alternative splicing in the third isoform, termed Shortest (TbeSts). Most Antarctic fishes belong to the Perciform suborder Notothenioidei (Cryonotothenioidea), which comprises five families (Nototheniidae, Harpagiferidae, Artedidraconidae, Bathydraconidae, Channichthyidae) and about 130 species of marine fishes found in the Southern Ocean, with a circum-Antarctic distribution, but also found in the more temperate coastal waters of the southern hemisphere [19]. Notothenioidei are among the most intensively studied lineages of marine fishes since they are a rare example of massive adaptive radiation driven by the same selective pressures (e.g. isolation and cooling) that may have led to the dramatic extinction of most fish fauna in the Southern Ocean [20][21][22]. The evolutionary success of Notothenioidei is marked by the acquisition of key adaptive features that enabled cold adaptation, such as the expression of antifreeze glycoproteins [23,24]. At the same time, Notothenioidei lost other traits universally shared by non-Antarctic metazoans, such as the inducible heat shock response [25] and, in the family Channichthyidae, hemoglobin [26,27].
Over the past 20 years, molecular and morphological studies allowed a revision of the phylogenetic relationships among notothenioid lineages, contributing to improve our knowledge about the adaptive radiation of these organisms [28][29][30][31][32]. Apart from the Antarctic Clade [33], three non-Antarctic lineages, distributed in proximity of the Southern Ocean, i.e. the southern regions of South America, around the Falkland Islands, Tristan da Cunha, New Zealand and south-eastern Australia, are currently recognized [20,34]. While the first one, Bovichtidae, includes six species, the two other families are monotypic and therefore include a single species, i.e. Pseudaphritis urvillii (family Pseudaphritioidae) and Eleginops maclovinus (family Eleginopsioidea). More recently, much attention has been dedicated to these notothenioid taxa, due to their early divergence from the polar lineage, which occurred before the climatic and geographic isolation of Antarctica, the drastic reduction in water temperature, and prior to the morpho-physiological diversification of cryonotothenioid species [35]. The study of the evolutionary history of Bovichtidae, based on mitochondrial and nuclear DNA molecular phylogeny, as well as on morphological and meristic characters, has been an essential factor for clarifying the process that drove the diversification between Cryonotothenioidea and their non-Antarctic relatives. Indeed, the evolutionary radiation of the genus Bovichtus and of the closely related cryonotothenioid species are characterized by a similar timeline, and therefore the patterns of diversification and extinction observed in these two lineages are expected to closely match each other [35]. Moreover, the revised positioning of E. maclovinus as the sister lineage of Cryonotothenioidea (it was previously considered as closely related to the nototheniid lineage Dissostichus) [30], provides another key information for the study of the evolution of these fishes. Taking into account the updated information about notothenioid phylogeny, the present work aims to extend the molecular analysis of IgT to the other notothenioid families and to solve the key question as to whether the features of the T. bernacchii IgT are unique to this species or shared by other Antarctic species. In order to track the evolutionary history of the Cτ2 exon and to pinpoint the timing of its partial loss, we obtained and comparatively investigated the IgT sequences of representatives from each of the five Antarctic notothenioid families and the three non-Antarctic lineages of Bovichtidae (Bovichtus diachantus and Cottoperca gobio), Eleginopsioidea (E. maclovinus) and Pseudaphritioidae (P. urvillii). The findings reported here bring further insights into the molecular evolution of the IgT gene in Antarctic fishes, marking the loss of the Cτ2 exon before the split between the Eleginopsioidea and Cryonotothenioidea lineages, and revealing early signatures of this event in the other early-branching non-Antarctic Notothenioidei.

2.1
The IgT cDNA sequence of Eleginops maclovinus provides new evidence about the origin of the loss of

Cτ2 exon in Antarctic species
Our investigations on the IgT cDNA sequences targeted several Antarctic species belonging to the Nototheniidae, Artetidraconidae, Bathydraconidae and Channichthyidae families (see the Materials and methods section for details), but also included the non-Antarctic species E. maclovinus due to its phylogenetic placement as a sister lineage of Cryonotothenioidea.
We obtained partial cDNA sequences, coding for the CH region of the IgT secreted form in all species analyzed, with the single exception of E. maclovinus, where the IgT membrane-bound form was obtained. The multiple sequence alignment highlighted the high conservation of the Cτ1 and Cτ4 exons in all species, as expected from previous publications (Figures 1 and S1). Mirroring the previously reported case of T. bernacchii, all Antarctic teleosts displayed a truncated Cτ2 exon, whose size ranged from 24 to 51 nt ( Figure 1, Table 1). obtained in a previous work [18], have been added for comparison. Cτ exon boundaries are reported above the alignment. Gaps are indicated by dashes. Below the alignment, identical nucleotides are marked with an asterisk, positions where only one sequence shows a different nucleotide are marked with a dot, positions differing in two nucleotides are marked with a colon. The duplicated 9-nt sequence at the beginning of the Cτ3 exon of the Hve2 transcript is underlined. Since several sequences varying in length at the 5' and/or at 3' were obtained from each species, only the region of each representative sequence that aligned over the same length has been shown here. Full alignments are provided in Figure S1.
The two variants cloned in E. maclovinus displayed a partially deleted Cτ2 exon (30 or 36 nt long in Ema1 and Ema2, respectively; see Table 1), revealing a Cτ2 exon structure similar to Antarctic Notothenioidei. On the other hand, we have previously shown that B. diacanthus, a non-Antarctic species more distantly related with Cryonotothenioidea, retains the entire Cτ2 exon (285 nt) [35] ( Table 1). Although Antarctic species and E. maclovinus shared the peculiar structure of Cτ2, their sequences differed due to several characteristic codon indels. In detail, the Cτ3 exon of all Antarctic species lacked one codon at positions 502, 598 and 617 (the latter was not missing in N. coriiceps), and two consecutive codons at position 562. On the other hand, a single codon insertion was found at position 526 in the Antarctic lineage ( Figure S1). Interestingly, all Cryonotothenioidea also presented four additional codons in the highly conserved Cτ4 exon at positions 715, 718, 778 and 784 ( Figure   S1). the Cτ1-Cτ2 and the Cτ2-Cτ3 introns). Based on the data reported in the previous paragraph, we used E. maclovinus as a reference for comparative analyses (Figures 2 and S2).
The two IgT genomic variants found in this sub-Antarctic species (Ema1 and Ema3) were characterized by a Cτ2 exon of variable size (30/36 nt) and matched those identified at the cDNA level. The Cτ1-Cτ2 intron of the isoform with the shortest Cτ2 exon (Ema1) displayed the insertion of a CAAACA sequence immediately before the splicing acceptor site of the Cτ2 exon ( Figure 2). On the other hand, the Cτ2-Cτ3 intron had identical length in both isoforms.
The two introns showed a significant size variation among Antarctic species. The Cτ1-Cτ2 intron ranged from 212 nt in N. coriiceps to 318 nt in T. bernacchii, which contained two 46-nt long repeated regions (Table 2, Figure 2). N. coriiceps and G. gibberifrons displayed an insertion of additional 15 nt at the 3' end of the intron, which was paired with the presence of the shortest Cτ2 exon out of all the species analyzed (24 nt, 8 aa). The size and structure of Cτ2-Cτ3 intron was much more conserved across Antarctic species, ranging from 115 nt in G. gibberifrons to 128 nt in H. velifer.
The short length of the G. gibberifrons intron was due to a 9-nt long deletion, which matched the position of a 7-nt indel evidenced by the multiple sequence alignment in E. maclovinus ( Figure 2). The alignment of the partial genomic sequences (Figures 3 and S3) hinted that the evolutionary process that led to the partial loss of the Cτ2 exon might have already started before the split between the Eleginopsioidea and Cryonotothenioidea lineages, as early as in the late Cretaceous.
Indeed, the gene of P. urvillii (Pseudaphritioidea) was characterized by early molecular signatures of erosion shared with the Antarctic species or in their sister taxa E. maclovinus. In particular, a few informative small deletions, matching the position of similar gaps in the sequence of E. maclovinus, were found in the Cτ1-Cτ2 intron, but not in the Cτ2-Cτ3 intron, which only included a few indels shared by all species, regardless of their position in the phylogeny of Notothenioidei ( Figure 3).
Unlike E. maclovinus, P. urvillii retained a complete Cτ2 exon, which only lacked the first three nt at its 5' end.
The sequences of the two species belonging to Bovichtidae, the most early-branch of the Notothenioidei lineage, also retained a complete Cτ2 exon (e.g. 282 nt -94 aa in C. gobio and 285 nt -95 aa in B. diacanthus, see Table 1) and displayed much larger introns than E. maclovinus (Table 2). However, the Cτ1-Cτ2 intron of C. gobio was also characterized by the presence of a few indels in similar positions to those observed in P. urvillii and E. maclovinus ( Figure 3).
The multiple sequence alignment of the Cτ2 exon of non-Antarctic Notothenioidei interestingly revealed that, despite the nearly complete deletion of the Cτ2 exon, E. maclovinus retained the six nucleotides found at the 3'end of its remnant nearly intact (with the single synonymous substitution TCG/TCT, see position 537 in Figure 3). This observation is in line with the high conservation of the canonical donor and acceptor splicing sites of the remnant Cτ2 exon observed in all species ( Figure   3).

The absence of repeats in the introns of the IgT gene rules out the involvement of transposable elements in the Cτ2 exon loss
As a further step in the investigation of the molecular mechanisms that might have led to the shortening of the Cτ2 exon, we investigated whether any repeated element could be identified in the Cτ1-Cτ2 and Cτ2-Cτ3 introns of the IgT gene from Antarctic nototheniods and in their sister lineage Eleginopsioidea. The activity of transposable elements (TEs) is well known to be associated with accelerated mutation, disrupting exons [36], altering splicing patterns [37], and shuffling the position of entire exons or of their parts by moving them to different genomic locations [38]. Hence, the presence of repeats could be indicative of the presence of active TEs, which may have possibly accelerated the evolution of the Cτ2 region. Our analyses revealed that neither the Cτ1-Cτ2 nor the Cτ2-Cτ3 introns contained traces of repeats in any of the species analyzed in this study.

The IgT gene sequence phylogeny is consistent with the phylogenetic relationships among Notothenioidei
In line with the observations provided above, we found that the molecular evolution of the Cτ1-Cτ2 and Cτ2-Cτ3 introns closely followed the currently accepted phylogenetic relationship among Notothenioidei. In detail, all the sequences from Cryonotothenioidea were placed in a highly supported monophyletic clade (100% posterior probability, Figure 4). As the sequence divergence between the different Antarctic species, which have been subject to a fast evolutionary radiation [39], was minimal, the topology of the Cryonotothenioidea subtree was characterized by very short branches. However, the two variants found in G. gibberifrons and H.
velifer were closely related (100% posterior probability), suggesting that both have been originated by species-specific gene duplications (the most likely hypothesis in G. gibberifrons, since one of the two variants is pseudogenic) or that they represent allelic variants.
As expected, the sequences from the Bovichtidae C. gobio and B. diacanthus were placed as outgroups in a monophyletic clade at the base of the notothenioid IgT tree, whereas the sequences from P. urvillii and E. maclovinus occupied, with high statistical support (100% posterior probability in both cases), intermediate positions. These were well consistent with the recently proposed position of Pseudaphritioidea and Eleginopsioidea [35]. Curiously, the two variants from E.
maclovinus shared closer homology with the Antarctic species than with P. urvillii and Bovichtidae, confirming the high relevance of this key species for the investigation of IgT evolution in Notothenioidei and further supporting the observation of shared indels, which may indicate a process of progressive loss of the Cτ1-Cτ2 intron.
Overall, the structure of this intron can be summarized by the presence of 5 distinct conserved sequence motifs ( Figure 5). Starting from the 5' end, the first 15 nt-long motif (named motif 5 in coriiceps, whereas the highly conserved motif 1 (50 nt-long) was found in all species and corresponds to the previously mentioned duplicated region found in T. bernacchii ( Figure 5). The last motif identified was the highly conserved 29 nt-long motif 4, which preceded the second repeat of motif 2.
The analysis of the nucleotide composition of the Cτ1-Cτ2 intron did not reveal any significant bias in Antarctic species (Figures 6 and S6), but at the same time it revealed an interesting trend in the Cτ2-Cτ3 intron. Indeed, in line with the placement of E. maclovinus as a sister group Cryonotothenioidea [30], the intron of this species had an AT content similar to the Antarctic species, which all showed a similar AT content (>70%) regardless of their length, and significantly higher than the other three non-Antarctic species (see Figure 6).  diacanthus (Bdi) and T. bernacchii (Tbe), previously obtained [18], and with that of G. aculeatus [40], the closest temperate species to the notothenioid species.
We investigated whether the IgT sequences of Antactic species were associated with the presence of conserved motifs that could be considered as possible "cold hallmarks", due to their in CH3 ) (Figure 7), none of these contained sites subject to purifying selection ( Figure 8).
We could detect, however, a few sites subject to significant purifying selection in the two exons flanking the Cτ2 exonic remnant. In detail, four and one negatively selected sites were detected in the Cτ1 and Cτ3 exons, respectively. The former exon included codons encoding highly conserved Leu, Asn and a Val residues (plus a Glu/Val residue found at the N-terminus of the domain, but just observed in three sequences), whereas the latter included a single codon encoding a Phe/Ile residue  Unlike temperate and non-Antarctic species, where sequons are evenly distributed along CH1 and CH3, in Cryonotothenioidea most glycosylation sites were found in CH3 ( Figure S5).
Although N. coriiceps and G. gibberifrons presented an Asn-Pro-Ser sequon in the CH2 remnant (also found in the CH4 domain of Artedidraconidae and Bathydraconidae), this is unlikely to be a real glycosylation site due to the proximity between a Pro and an Asn residue, which is expected to make the Asn residue inaccessible [41].

Discussion
The immunoglobulins of Antarctic fishes have been fascinating us since the early discovery of unforeseen features of IgM from Cryonotothenioidea [42][43][44]. For several years, our studies have been mostly focused on IgM, an ancient Ig isotype that first appeared in jawed fish along with the emergence of an adaptive immune system [45]. However, we recently moved our attention to the study of the heavy chain gene of IgT, a fish-specific Ig isotype, whose discovery revealed the origins of the most ancient Ig specialized in mucosal immune response. We disclosed that the gene of Antarctic species T. bernacchii, unlike the early-branching non-Antarctic notothenioid species B.
diacanthus and most species living in temperate environments, displayed an unusual truncated Cτ2 exon, which only encoded a short remnant of the CH2 domain [18]. This finding was the starting point of the present work, which extends our molecular investigations to other Antarctic species and to the early-branching non-Antartic notothenioid lineages, in the attempt to pinpoint the origins of this partial exon loss event.
The groundwork for placing down the first piece of this evolutionary puzzle was provided by the observation that the Cτ2 exon was nearly completely missing also in E. maclovinus, a non-Antarctic species which belongs to Eleginopsioidea, the sister group of Cryonotothenioidea.
This allowed us to move backwards through the phylogeny of Notothenioidei, characterizing the IgT sequence of C. gobio, belonging to the most basal group of non-Antactic Notothenioidei, i.e.
Bovichtidae. This species showed a nearly complete Cτ2 exon, except for a small deletion located at its 3'end of the exon, which matched a similar indel carried by E. maclovinus. Moreover, C. gobio also displayed a shorter Cτ1-Cτ2 intron than its close relative B. diacanthus, with a few deletions shared with E. maclovinus.
The cornerpiece of the puzzle was provided by the analysis of the IgT sequence of P. urvillii, be viewed as an ancestral characteristic of ectothermic vertebrates [46]. An integrated revision of the genomes readily available at present for many notothenioid species may provide a useful framework for assessing this issue. These observations are in line with key role that introns cover in the dynamic process of genome evolution [47]. The role of introns in the adaptation to varying environmental pressure may have been particularly relevant in teleosts, where these elements are found in higher number and usually have a shorter length than other vertebrates, as a consequence of the teleost-specific whole-genome duplication event [48,49]. We have previously revealed, as a key example of this potential for adaptation, the peculiar rearrangement of the exon/intron architecture of the region encoding the C-terminal Extracellular Membrane-Proximal Domain in the IgM heavy chain gene locus of Antarctic fish [50].
Although Transposable Elements (TEs), found in 20-60% of introns in vertebrate genomes, can also provide a significant contribution to the modification of gene architecture [51,52], we could not find any active TEs in the two introns flanking the Cτ2 exon, neither in Antarctic, nor in non-Antarctic species. However, we cannot exclude the possibility that TEs that were lost along evolution and that are not detectable anymore have played a role in the process that led to the loss of the Cτ2 exon.
Most certainly, the reconstruction of this scenario suffers from missing data and significant "evolutionary gaps" due to the high phylogenetic distance between the genera we took into account [35]. In any case, the IgT sequence features outlined above for non-Antarctic species are "molecular fossils" that can provide useful information to infer the stepwise Cτ2 exon loss that occurred in the Antarctic lineage.
The evolutionary dynamics of exon-intron architecture are another significant aspect of genome adaptation [53,54], as they accommodate a high variation of intron and exon size but require an extreme conservation of the splicing site motifs located at their boundaries [55]. We show that the Cτ2 exon remnant of all Antarctic species, despite its short length, retains both functional acceptor (5') and donor (3') splice sites. While the conservation of the donor splice site can be explained by the preservation of the 6 nucleotides at the 3'end of this exon, the entire 5'end of Cτ2 was deleted in Antarctic species, along with the acceptor splicing site. We speculate that the correct splicing might still occur in this region thanks to the presence of an AGGCAA motif, which nearly matches one of the five predicted splicing regulatory hexamers of mammals [54].
The recurrent reorganization of Ig gene loci during vertebrate evolution has often led to the generation of multiple functional gene copies and pseudogenes [56]. A similar evolutionary process might have targeted Antarctic IgT genes, thereby generating novel sequence variants that could have granted the acquisition of new structural arrangements and functions, favorable for the adaptation to the polar environment, as in the previously reported case of T. bernacchii. suggested by (i) its high solvent exposure predicted by 3D molecular modeling ( Figure 9); (ii) the abundance of amino acid residues typically found in hinge regions (i.e. proline, glycine and cysteine), recalling in particular the human IgA1 hinge region, which is also of similar length.

Most of the variants reported in this work had a length similar to the variant S of T. bernacchii
The finding that the Cτ2 exon remnant, despite its short length, contained two sites subject to diversifying selection was intriguing. While the first one was found in a region characterized by indels in some species, the second one, encoding Pro/Ser/Thr, was adjacent to a highly conserved cysteine residue. This observation will undoubtedly lead to further investigations aimed at clarifying the functional role of these residues in the context of cold adaptation, which presently remains unknown. Cysteine pairs involved in the formation of disulfide bonds are highly conserved relative to unpaired cysteines and to other amino acids [58]. Consequently, whenever disulfide bonds are permanently consolidated in proteins, the cysteine residues involved in the formation of such bonds rarely vary. Since the remnant portion of the CH2 domain just maintained the second of the two canonical cysteines, we speculate that the high conservation of this residue might be linked with a function in: (i) bridging the two heavy chains in the monomer; (ii) keeping the monomer units covalently linked in the multimeric form, creating the so-called "redox forms" observed for other Ig isotypes.
In general, variations in disulfide connectivity allow a higher degree of polymerization, influencing the structure, stability and effector functions of Ig molecules [59], as it typically happens in the case of IgM [60]. In light of these observations, it is tempting to speculate that the remnant CH2 domain may provide an advantage for the activity of this Ig isotype under the thermodynamically unfavorable cold conditions of the Antarctica.
Although little information is available concerning the aggregation status of fish IgT, biochemical studies carried out in the rainbow trout have revealed that they are present in the serum as a monomer and that, unlike IgM, the multimeric complexes found in the mucus are non-covalently linked [61]. Considering that teleosts lack the J chain, essential for the formation of Ig polymers, the principles that drive their association in multimers are presently unknown.
Teleost mucosal Igs can associate with the polymeric Ig receptor (pIgR), which is also present in T. bernacchii (unpublished data) and possibly in other Antarctic fish, and enables their secretion into the gut lumen, similar to mammalian IgA and IgM [61,62]. Antarctic IgT contain, at the carboxy terminus of the secretory tail, a sequence motif similar to the one found in other teleosts, which may be involved in polymerization.
A third factor that might allow IgT polymerization relies on glycosylation, even though the role of glycan moieties in teleost mucosa-associated Igs still needs to be clarified. The IgT of Antarctic species contain a high number of sequons consistent with glycosylation sites. In particular, the Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 27 November 2020 doi:10.20944/preprints202011.0685.v1 presence of two such sites in the C-terminal secretory tail of the Antarctic IgT may suggest a role in the assembly of Ig complexes, as they match the position of conserved glycosylation sites found in mammalian IgM and involved in the polymerization of this Ig isotype [63].

Biological samples
The biological material was collected from a set of species representing the Antarctic families  Table S1.  Table S4.

Additional sequences obtained from public repositories
Sequence data concerning the IgT gene and cDNA sequences from T. bernacchii (family Nototheniidae) and from B. diacanthus (family Bovichtidae) were retrieved from a previous study carried out by our research team [18]. The genomic DNA and predicted cDNA sequences of the IgT genes of Notothenia coriiceps (family Nototheniidae) and C. gobio were retrieved from publicly available annotated genome assemblies. In detail, the N. coriiceps genome refers to the study by Shin et al. [28] whereas the genome of C. gobio (v. fCotCob3.1) has been recently released within the frame of the Vertebrate Genome Project [64].

Computational analysis
Sequencing chromatograms were visualized using the program FinchTV (version 1.3.0). The nucleotide sequences obtained were verified by sequence similarity searches against the GenBank database, using the BLAST program [65]. Amino acid sequences were deduced from nucleotide sequences using the ExPASy Translate Tool tool. The amino acid composition was analysed using

Association between IgT genes and repeated elements
The reference genomes assembly of C. gobio [64] was analyzed with RepeatScout v.1.05 [69] to generate a species-specific repeat library. The IgT genomic DNA sequences obtained in this study, as well as those identified in the genome of C. gobio, were analyzed with RepeatMasker v.4.09 [70], with particular attention to the Cτ1-Cτ2 and Cτ2-Cτ3 intronic regions. All sequences were screened for the presence of repeats against the Dfam v.3.1 [71] library of known repeats found in the genomes of Actinopterygii. The IgT gene of C. gobio was subjected to an additional round of screening against the custom species-specific repeat libraries generated as described above.

Molecular evolution of the Cτ1-Cτ2 and Cτ2-Cτ3 introns
We investigated the evolutionary history of the IgT genes identified in Notothenioidei, with particular focus on the genomic region subjected to the highest molecular diversity, i.e. the region spanning the Cτ1-Cτ2 and Cτ2-Cτ3 introns. Due to the truncation of the Cτ2 exon in Cryonotothenioidea (see the results section), this region was disregarded. The nucleotide sequences of the two introns were separately aligned with MUSCLE v.3.8.31 [72] The multiple sequence alignments were manually refined and processed with GBLOCKS v.0.91b [73] to remove phylogenetically uninformative positions. The two sequence blocks were concatenated and tested with ModelTest-NG v.0.1.3 [74] to identify the best-fitting model of molecular evolution. This was found to be the GTR+I model (Generalized time-reversible, with a proportion of invariable sites) [75], based on the corrected Akaike information criterion [76]. The multiple sequence alignment file was used as an input for Bayesian phylogenetic inference, carried out with MrBayes v.3.2.7a [77], run for one million generations and two parallel Markov Chain Monte Carlo (MCMC) analyses. The convergence of the two independent analyses was checked with Tracer [78], by evaluating that all the estimated parameters reached an ESS value higher than 200.

Selection analyses
The Ig Cτ cDNA sequences of the available species from Cryonotothenioidea, either determined by cloning or inferred from genomic DNA, were aligned with a strategy aimed at preserving the integrity of codon triplets. This was achieved by aligning the translated amino acid sequences with MUSCLE v.3.8.31 [72] within the MEGAX environment [79] and back-translating the aligned sequences to the original gapped coding nucleotide sequence.
The multiple sequence analysis was subjected to tests aimed at detecting signatures of selection with the DataMonkey adaptive evolution platform [80]. Sites evolving under pervasive positive and negative selection were detected with FEL (Fixed Effect Likelihood) [81] and those with evidence of episodic positive selection were identified with MEME (Mixed EffectsModel of Evolution) [82] using default p-value thresholds.

Identification of short conserved sequence motifs in Antarctic species
The unaligned IgT protein sequences of Cryonotothenioidea were analyzed to detect the presence of short ungapped conserved amino acid motifs of 3-8 aa length with MEME (Multiple Em for Motif Elicitation) [83] within the MEME suite v.5.0.1 environment [84]. The specific association of the detected motifs with Cryonotothenioidea was subsequently tested by assessing their absence in the IgT sequences of non-Antarctic Notothenioidei and temperate Perciformes. Finally, the relevance of the identified motifs in the context of evolution in the Antarctic environment was evaluated by inspecting their overlap with negatively selected sites, identified as explained in the previous section.
The Cτ1-Cτ2 and Cτ2-Cτ3 introns of all the available Notothenioidei sequences were similarly screened, looking for conserved nucleotide motifs with size comprised between 6 and 50 base pairs, allowing any number of motif repeats for each sequence.

Conclusions
The many challenges notothenioid fishes faced during their evolutionary history triggered a wave of genomic changes that led to a number of peculiar features that might have been preserved as adaptive traits. The development of an IgT molecule nearly completely devoid of a CH domain might be just another addition to the list of the highly successful strategies these organisms have used to adapt to this challenging environment. The evolutionary scenario we propose involves a gradual process, which has shaped the IgH gene locus over a long period of time, through the action of contrasting evolutionary forces, contributing to the generation of an Ig molecule with a unique architecture among vertebrates.

Acknowledgments
The authors wish to thank Dr Ennio Cocca (IBBR, CNR, Naples, Italy), for providing tissue samples of E. maclovinus, C. rastrospinosus, and C. aceratus specimens; Drs Daniela Giordano, and Cinzia Verde (IBBR, CNR, Naples, Italy) for providing P. urvillii and G. gibberifrons blood samples. The authors are grateful to Prof.
Alberto Pallavicini (University of Trieste, Italy) for his critical comments on the manuscript.
In memory of Guido di Prisco, for his scientific career entirely devoted to unveil the mystery of evolutionary adaptations for life in Antarctica.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.