Introduction

Most marine teleosts are unable to inhabit ice-laden sea waters characteristic of polar and sub-polar oceans because the temperature of the water (−1.9°C) can be a full degree lower than the freezing point of their body fluids (−0.7 to −0.9°C) [1]. In contrast such environmental conditions pose no risk to most invertebrates as their freezing points are usually the same as that of sea water [2]. Teleost fish that can survive and thrive in such environments do so by producing antifreeze proteins (AFPs) or glycoproteins (AFGPs) that bind to nascent ice crystals within their body fluids, thereby preventing their further growth that would ultimately result in death [3–7]. The difference between the melting point and non-equilibrium freezing point is defined as thermal hysteresis (TH) and is a measure of AFP activity [5].

To date, three non-homologous, physiologically functional groups of AFPs (types I-III), as well as the AFGPs, have been described in a variety of fish taxa [8–10]. Both type II and type III AFP, are globular and non-repetitive, with the type II AFP gene having been horizontally transferred between three fish orders (Osmeriformes, Perciformes, and Clupeiformes) that diverged over 200 Ma ago [11, 12]. In contrast, both AFGP and type I AFP are repetitive proteins. AFGPs consist of variable numbers of glycosylated tripeptide repeats and arose by convergent evolution in Antarctic Notothenioids (Order Perciformes) and the unrelated northern cods (Order Gadiformes) from two different progenitor sequences [13–16]. The repetitive alanine-rich α-helical type I AFPs also arose by convergence, within four families from three orders: Pleuronectiformes, Labriformes and Perciformes [17]. All of the AFP-producing fish orders mentioned above diverged prior to the Eocene, which began with polar oceans that were ice-free during the Paleocene–Eocene thermal maximum at 55 Ma [18, 19]. Therefore, the impetus behind the evolution of these unique ice binding structures was the subsequent occurrence of sea ice following the onset of global cooling and glaciation [20]. These are clear examples of convergent protein evolution to a common function [21]. The occurrence of lateral gene transfer [12], and of convergence to highly similar repeat sequences [13, 17] adds a level of evolutionary complexity that is absent from most gene families.

In contrast to the other fish AFPs and AFGPs, type III AFP is restricted to a single taxonomic group, order Perciformes, infraorder Zoarcales (previously suborder Zoarcoidei), that diverged approximately 50 Ma ago [19] (Fig 1). This ~7 kDa protein arose from the C-terminal domain of sialic acid synthase (SAS) following a gene duplication [22–24]. The first type III AFPs were isolated from ocean pout (Zoarces americanus, family Zoarcidae) [25, 26]. They are found as mixtures of SP-Sephadex- (SP) and QAE-Sephadex-(QAE) binding isoforms that are only about 55% identical. Both of these isoform sets have the capacity to bind to ice, but only a subset of QAE isoforms are able to completely halt ice crystal growth [27, 28]. Interestingly, the SP isoforms can be made fully active in stopping ice growth if as little as 1% of the QAE isoform is included with them [28].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Detailed relationships among species of the infraorder Zoarcales discussed in this study.

Divergence times were obtained from the following studies; Gasterosteiformes and Zoarcales, from 21 loci from a broad range of teleost fishes [19], Bathymasteridae, from 10 nuclear genes [88] and within Zoarcales [42] or Anarhichadidae [89] using fewer loci. A double asterisk denotes that the fish from these studies was the same as the species in this study, whereas a single asterisk denotes that the fish was a different species within the same genus. The 95% posterior density intervals for the branch points are shown by grey bars. Species names are coloured in a rainbow pattern from red to purple from earliest to latest time of divergence of each family. The location of the Cryptacanthodidae varies between studies (denoted with a dotted line) [19, 46] and some Stichaeidae cluster with Pholidae [44–46]. Information on Lycodinae and radiated shanny was limited. Therefore, their relationship to the others was determined using the COX gene with Alaskan ronquil as the outgroup (S2 Fig), with divergence times estimated from the nearest calibrated node. The nodes at which there is the earliest evidence for the presence (+) or absence (-) of SP or QAE isoforms are indicated, as well as branches along which gene duplications have occurred (SP↑). The number of known Zoarcales species and their range (green background for northern, pink for southern) was determined from FishBase [72]. The average temperature of the deep ocean, relative to the present day, as well as the temporal extent of northern (N) and southern (S) glaciation, was adapted from Fig 1 of [90].

https://doi.org/10.1371/journal.pone.0243273.g001

Fishes from the infraorder Zoarcales have a global distribution and are the only group of fishes that have been demonstrated to have AFP-producing species that have conquered the icy waters at both poles (Fig 1). Five families, four of which are restricted to Northern waters, were shown by early Southern blotting studies to possess numerous type III AFP genes: Zoarcidae [ocean pout (Z. americanus)]; Anarhichadidae [spotted wolffish (Anarhichas minor) and Atlantic wolffish (A. lupus)]; Pholidae [rock gunnel (Pholis gunnellus)]; Stichaeidae [radiated shanny (Ulvaria subbifurcata)] [29–31]; Cryptacanthodidae [wrymouth (Cryptacanthodes maculatus)] [32, unpublished data]. However, AFP sequences have only been obtained from sequences within Zoarcidae and Anarhichadidae [25–27, 29, 30, 33, 34]. Most of these species inhabit the shallow inshore waters of Newfoundland, where they are frequently exposed to subzero temperatures and ice during the winter months [35].

Type III AFPs have also been characterized from two species that reside in the frigid Southern Ocean around Antarctica. Both Lycodichthys dearborni and Pachycara brachycephalum are commonly called Antarctic eelpouts, so to differentiate them, we will refer to L. dearborni as Antarctic eelpout and P. brachycephalum by scientific name only. The AFP complement of the Antarctic eelpout has been studied through a combination of protein, cDNA and genomic DNA sequencing (yielding over 20 gene sequences). This species produces both monomers and tandemers consisting of two or more linked AFP domains [23, 36–40], whereas the northern species studied produce only monomers. Interestingly, there is no evidence of tandemers in the second Antarctic species examined, P. brachycephalum [39, 41].

Given the diverse evolutionary history of fish AFPs, we examined in this study several questions about type III AFPs. The first was to establish if this AFP has spread through the Zoarcales by direct descent as opposed to parallel evolution. By mining unique sequences from various databases and transcriptome studies, and through targeted sequencing projects on several species, we can confidently say that all known type III AFPs are related by descent. Secondly, we wanted to date the origin of the type III AFP, so we successfully expanded the search for AFP sequences in other Zoarcales families out to the Alaskan ronquil, Bathymaster caeruleofasciatus, which diverged around 18 Ma [42] (Fig 1). Finally, having established here that all type III AFPs are related by descent, we addressed the timing of the colonization of Antarctic waters by zoarcids. The most plausible explanation is one of recent migration of a founder species from the north that transitioned through the cold ocean depths linking the two poles.

Results

Type III AFPs arose early in the infraorder Zoarcales

To trace the evolutionary history of type III AFP and date its origin, we set out to expand the number of known sequences from the three Zoarcales species mentioned above (Alaskan ronquil, radiated shanny and rock gunnel that are found in two lineages that diverged early in this clade. Full-length cDNA sequences encoding both QAE and SP isoforms were obtained from rock gunnel and radiated shanny using primers based on known sequence (S1 Table) and a combination of RLM-RACE and RT-PCR. When the protein sequences of these two species are aligned (Fig 2, S3 Fig), the SP isoforms are 72% identical while the QAE isoforms are 79% identical. This identity drops to ~60% when SP and QAE isoforms are compared. The accession numbers and characteristics of these sequences are given in S2 Table. QAE and SP isoforms were originally categorized by their ability to bind to positively-charged quaternary aminoethyl (QAE) or negatively-charged sulfopropyl (SP) resins at neutral pH [26]. However, the rock gunnel-Q1 sequence has a predicted isoelectric point at pH 9.5, more like that of SP isoforms. Therefore, sequence similarity is a better way to categorize type III AFPs. When the cDNA sequences were aligned with those of other fish (S4 Fig) and used to generate a phylogenetic tree (Fig 3), the SP sequences clustered together in this phylogeny, as did the QAE sequences, consistent with the relatedness of these two species (Fig 1). However, the first exon, which encodes the signal peptide, was identical between the rock gunnel SP and QAE isoforms. Phylogenetic analysis of this short exon (S5 Fig), using the homologous region of SAS, which does not encode a signal peptide [23], to root the tree, shows that for all other sequences, QAE sequences cluster, as do SP sequences. In contrast, the rock gunnel and radiated shanny exons cluster outside of these two groups. This suggests that exon shuffling may have occurred in the common ancestor of these two species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Alignment of a representative selection of known type III AFP sequences.

Sequences are named using the common name for each species except for P. brachycephalum. Names are colored as in Fig 1 with an asterisk indicating the sequences were obtained by PCR in this study (S3 Table) and with a dagger for those assembled from the SRA database (S4 Table). They are numbered consecutively by species within the QAE (Q) or SP (S) group as they appear in S3 Fig. Residues conserved in all isoforms (sasA and B) are indicated with asterisks, variations characteristic of the QAE and SP groups are highlighted cyan and yellow respectively, with other variations in grey. Shared differences in the signal peptides of radiated shanny and rock gunnel are highlighted red. Black and red boxes show residues on the pyramidal- and prism-plane binding surfaces respectively [79]. Inward-facing residues (i) are indicated along the top. Residues unique to SAS are highlighted black with other differences within this group highlighted purple. Signal peptides are in lower case font. Dashes indicate gaps (internal) or incomplete sequence (at termini). The complete protein and nucleotide alignments are shown in S3 and S4 Figs, respectively.

https://doi.org/10.1371/journal.pone.0243273.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Phylogenetic and functional comparison of type III AFPs.

A) A maximum-likelihood phylogenetic tree of the nucleotide sequences (S4 Fig) of the subset of type III AFP sequences shown in Fig 2. Cyan and yellow backing denotes QAE and SP isoforms respectively and bootstrap values (percent) are indicated at most nodes. The scale bar represents an average of 0.1 or 1 changes per site for solid and dashed lines respectively. B) and C) Representative structure of a QAE (PDB:4UR4) and SP isoform (PDB:4UR6) respectively [53] with the pyramidal and prism ice-binding surfaces colored orange and cyan respectively [78]. D) and E) Diagram of the fluorescent ice-plane affinity of a QAE and SP isoform respectively, adapted from previously published images [79]. F) and G) Ice crystals in the presence of 0.1 mM of a fully-active QAE isoform (M1.1, [91]) and 0.46 mM of an SP isoform (notched-fin eelpout-S5, S3 Fig [27]) respectively. Samples were cooled at a rate of 0.01°C/6 sec for one min then held for the indicated times at 0.1°C below the melting point. The scale bar represents 10 μm.

https://doi.org/10.1371/journal.pone.0243273.g003

Northern blotting was performed using various tissues from both the rock gunnel and the radiated shanny (S6 Fig) and the transcripts were most abundant in the liver. Moderate expression was found in skin, gill and stomach, indicating that that the liver-dominant expression originally observed in Zoarcidae [47] and Anarhichadidae [29] is also a feature that arose early in Zoarcales. A partial type III AFP sequence from an even more divergent species, the Alaskan ronquil (Fig 1), was amplified from genomic DNA in two overlapping fragments using semi-nested PCR (see Supplementary Methods). This single sequence encodes a QAE isoform, Alaskan ronquil-Q1 (Fig 2). The gene structure is consistent with other type III genes. The 174 bp intron (not shown) lies close to the expected location, but both splice junctions are shifted leftward by three bp (S4 Fig). This sequence is most similar to Atlantic wolffish-Q2. The identities between the protein sequences, as well as between the coding sequences and the single intron, were between 94 and 95%. The 3' UTR is up to 97% identical to other type III sequences. SP sequences were not recovered, but their existence cannot be ruled out as the DNA amount and quality from the museum specimen was low and/or the primers used may not match Alaskan ronquil sequences. The presence of this sequence in the Alaskan ronquil pushes back the origin of type III AFP to approximately 18 Ma.

SP sequences cluster along family lines

There are two main groups within the SP cluster (S8 Fig), in which sequences from Anarhichadidae (red dashed box) cluster separately from those within Zoarcinae (blue dashed box). The sequences from radiated shanny and rock gunnel also cluster. In contrast, the QAE sequences that are known show a much weaker association by family (Fig 4). This suggests that the common ancestor of these families may have possessed a larger number of QAE sequences than SP sequences. Alternatively, the SP genes may have undergone rounds of expansion and contraction more frequently than QAE genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Phylogenetic and geographical distribution of QAE isoforms.

A) Phylogenetic tree of the complete QAE subgroup from S4 Fig. Sequences are indicated as in Fig 2 and sequential tandemers from Antarctic eelpout are labelled alphabetically (e.g. Antarctic-eelpout-Q3a, Antarctic-eelpout-Q3b). Sequences from northern fish are on a green background while those from southern fish are on pink. The scale bar represents an average of 0.02 changes per site and bootstrap values (percent) are indicated at most nodes. B) A newly-discovered Pachycara sp., found during the 2016 NOAA Okeanos expedition at the Mariana Trench, 18°27ʹ N; 147°50ʹ E, in 1.5°C waters below 4000 m (https://service.ncddc.noaa.gov/rdn/oer-rov-cruises/ex1605l3/#tab-20). C) Diagram of a cold-water migration route through a cross-section of the oceans from the Arctic to the Anarctic. The deep water passage shown by the dotted arrows through the tropics avoids warmer surface waters (red).

https://doi.org/10.1371/journal.pone.0243273.g004

Discussion

Type III AFP, which is found in fishes from a single branch of the taxonomic tree, may have allowed the Zoarcales to diversify and spread during the last 18 Ma, during a time of global cooling (Fig 1). Like most other fish AFPs, these are also present in multi-gene families [56]. We add to the body of evidence that this gene family has undergone many changes during this period, in which gene losses, gene duplications and other mutational events have occurred. The impetus for these changes is likely related both to the changing climate and to the migration patterns of the various species examined.

The Paleocene and early Eocene, from 65 to 45 Ma, was much warmer than today, with an ocean that was ice-free [18]. The presence of cold-intolerant tropical vegetation along the Antarctic coastline during the early Eocene [57] and deep ocean temperatures up to 14°C higher than today [18], support this assertion. The Zoarcales diverged from the Gasterosteiformes (sticklebacks and relatives) around this time [19], so the absence of type III AFPs outside of Zoarcales is not surprising, given the common ancestor of these two groups would have had no need for an AFP.

Once the Earth began to cool and ice was again present in the oceans, fish that acquired AFPs would be able to exploit “freeze-risk ecozones” with their abundance of invertebrate prey [29, 58]. This scenario is considered the impetus for the evolution and diversification of the four known AF(G)P types in diverse fish taxa [20]. An excellent example of ecozone exploitation is provided by the Notothenioids, which became the dominant taxon in Antarctic waters after acquiring AFGPs [59, 60].

The situation in the Northern Hemisphere is more complex as species with all four functional AFP types are found here [20]. In addition, the climate history of the Arctic is not as well understood. While Antarctic glaciation began ~35 Ma, the Arctic remained warmer for far longer with widespread glaciation apparently only occurring within the last 3.5 Ma [61, 62]. However, proxy evidence such as diatoms assemblages and ice-rafted debris suggest that ephemeral Arctic sea ice formed far earlier than previously thought, prior to 40 Ma [63, 64].

It is generally accepted that the infraorder Zoarcales originated in the Northern Pacific [65–68]. Therefore, our discovery of type III AFPs in two northern Zoarcales species that diverged over 10 Ma (radiated shanny, rock gunnel) and the Alaskan ronquil that diverged ~ 18.4 Ma (based on a time-calibrated phylogeny [42]), lends credence to the hypothesis that sea-ice was abundant in the North Pacific, well in advance of the opening of the Bering Strait [69–71] and many Ma before widespread glaciation occurred. It is likely that these groups remained in the Northern Pacific Ocean as nearly all species from the three families to which these fish belong (Bathymasteridae, Pholidae and Stichaeidae) still reside in this area [72]. It may be that the ancestors of the radiated shanny and rock gunnel migrated through the Bering Sea once the Bering Strait opened ~5 Ma [69–71], as the Bering Sea is thought to be the location where the Zoarcales underwent a major radiation [67]. The AFPs of these fish would have allowed them to survive in these icy northern waters. Species from these three groups never crossed the equator however [72], but as the surface waters in the tropical Pacific Ocean were only 1 to 2°C cooler during the last glacial maximum than they are today [73], they may have acted as a barrier to the spread of species adapted to survive in colder waters.

The SP and QAE types arose early within the Zoarcales lineage as they are found within all northern fishes for which multiple sequences were obtained [Zoarcidae [26], Anarhichadidae [29, 30], Pholidae and Stichaeidae (this study)]. The progenitor is under strong negative selection as the entire SAS sequence is highly conserved, as shown by the near-identity (96%) of SAS-B from a northern wolf eel (family Anarhichadidae) to that from Antarctic eelpout (family Zoarcidae). The AFPs are far more variable and show a high rate of mutation on surfaces that are not involved in binding to the pyramidal ice plane. Despite this, all of the AFP sequences, whether they belong to the QAE or SP groups, are far more similar to each other than they are to SAS. Additionally type III AFP is only found in one infraorder (Zoarcales), whereas the three other AFP types that are known to have arisen by convergence or lateral gene transfer are found in fish from different orders [12, 14–17]. Taken together, it is very unlikely that type III AFP arose more than once from SAS (convergence, specifically parallelism). Instead, a single AFP likely arose early during the diversification of suborder Zoarcales and following one or more duplication events, gave rise to the SP and QAE isoforms that were transmitted by vertical descent to various families within Zoarcales as they arose.

As fishes from all but the most recently evolved family within Zoarcales are restricted to northern waters (Fishbase, [72]), the type III AFP clearly arose in the Northern Hemisphere. The eelpouts (Zoarcidae) originated around 10 Ma and are one of only a few families of fish found at both poles [74]. How then did some Zoarcids move from the far north to the Antarctic Ocean? It seems unlikely that cold-water fishes would migrate into warmer tropical waters and then back into cold Antarctic waters, all while retaining AFPs. They would need to first adapt to warmer water, where they would face competition from extant warm-water species, and then adapt back to colder water on the way to Antarctica. This would have been the case even during the coldest periods of the ice ages, because although the extent of the warmer waters was reduced, both the tropical Atlantic and Pacific waters were only a couple of degrees cooler than they are today [73]. However, there is another possible route, in both the Pacific and Atlantic basins, that does not require incursion into warm waters. Cold-adapted species could move from pole to pole through deep, perpetually cold waters (Fig 4). Indeed, Zoarcids are the predominant species near hydrothermal vents where the temperature a short distance away is typically around 2°C, and Pachycara spp. have been found at depths exceeding 3 km in both the Atlantic and Pacific basins (Fig 4) [75, 76]. Interestingly, the majority of the Zoarcids that have been found near the equator are demersal (bottom dwelling) fish that have been recovered from depths of 500 m or more [72]. Unfortunately, the present-day distribution of the Antarctic eelpout and P. brachycephalum does not provide further clues as to their migration route as the Southern Ocean lacks land barriers.

Another possible scenario that could explain the presence of type III AFPs in southern Zoarcids is that the fish that migrated south did not have AFPs. Instead, they could have evolved anew from SAS once they encountered the icy Southern Ocean (parallelism). Work by Deng et al. (2010) in Antarctic eelpout showed that it was the SASb gene that was duplicated, with the 3ʹ exon encoding all but the signal peptide of the AFP [23]. The comparison between the C-terminal domain of SASa and SASb from wolf-eel, a northern fish within the same family as the spotted and Atlantic wolffishes, to the corresponding sequences from Antarctic eelpout, casts doubt on this hypothesis. If the AFPs arose independently from SAS progenitors in the two hemispheres (parallelism), the extant AFPs should show more similarity to their conspecific progenitors than to each other. However, the phylogenetic trees indicate the exact opposite (Fig 3). Additionally, there would be no reason for the Antarctic sequences to be most closely related to the single sequence known from Canadian eelpout, which is the closest relative to the Antarctic species in our study, with these two lineages having diverged less than 10 Ma (Fig 1). However, the Canadian eelpout sequence is 87% identical to Antarctic eelpout Q4 and clusters with the Antarctic sequences in the phylogenetic tree (S7 Fig). Furthermore, all of the AFPs clustered together near the end of a long branch (Fig 3), suggesting that the AFPs did not begin duplicating until the primordial gene had diverged significantly from SAS. Although it is plausible that the residues on the ice-binding surfaces could be similar due to convergence, the similarities between the variable surface residues away from these sites is strongly suggestive of homology. Therefore, the AFPs within the Antarctic species are clearly homologs of those found in northern fishes.

An alternative hypothesis to parallelism is that the population that migrated south, long after the onset of Antarctic glaciation, may have lost all but one or a few nearly-identical AFP genes during the journey as all of the Antarctic sequences form a single cluster within the QAE portion of the phylogeny. While it is difficult to prove the complete absence of SP isoforms, the fact that none were recovered from serum or the transcriptome of P. brachycephalum [41, 50] or following screening of EST and BAC libraries from Antarctic eelpout [39, 40] strongly suggests they are absent. Unfortunately, genome sequencing has not proven useful for characterizing this multi-copy gene family, as shown by the failure of the AFPs of wolf eel to properly assemble. Therefore, the sequencing of individual BACs likely provides a clearer picture of the gene complement in Antarctic eelpout [23, 39]. Conversely, it is possible that other species of southern Zoarcids did retain SP isoforms and that there may have been more than one species that migrated into southern waters. The plasticity of the AFP gene family, both with regards to gene number and gene organization, has been clearly demonstrated. For example, the number of AFP genes, even within a single species, can change dramatically as was seen with ocean pout living at different latitudes [26]. Additionally, the organization can differ as the AFP genes are found in tandem repeats in ocean pout [26], as inverted pairs within tandem repeats in Atlantic wolffish [30] and in tandem repeats containing some tandemers at a single locus in Antarctic eelpout [23, 39]. rDNA genes are found in similar arrays and they are known to undergo rapid changes in copy number (reviewed in [77]). It should be noted that there is no well-documented advantage of AFPs to fish other than to protect them from internal ice growth. Thus, there is no need for AFPs in cold deep water where ice crystals are absent. Therefore, the phylogenetic analysis supports the hypothesis that gene duplication/amplification of the remaining QAE type gene(s) in the newly arrived migrants would have allowed these fish to survive in the icy Antarctic waters.

Further evidence for the plasticity of the AFP gene family is provided by the additional sequences we obtained from ocean pout, and by analyzing the transcriptome of the viviparous eelpout [49]. A variety of divergent QAE isoforms have been retained in both of these eelpouts as the QAE isoforms of these two species form a number of clusters. However, the SP isoforms of viviparous eelpout cluster together with those of the notched-fin eelpout but separately from those of ocean pout. This suggests that gene losses and duplications may have occurred frequently, particularly within the SP group, within the last 10 Ma. Still, both types have been retained. Many isoforms (all SP and some QAE) have been shown to impart hexagonal shaping to ice without preventing growth, due to structural differences in the ice binding site that prevents binding to the primary prism plane of ice [78, 79]. This has led some to suggest that these AFPs should be reclassified as "ice growth modifiers" [28, 80]. This limitation is overcome by the addition of minor quantities of a "fully active" QAE isoform [27, 28]. The isoforms appear to show synergism as the activity of the mixture is a little higher than that obtained with the QAE isoform alone [28]. Synergism has also been reported with AFGPs [81, 82] and beetle AFPs [83]. Therefore, the retention of multiple isoform types is likely advantageous for type III-producing fishes.

The two Antarctic species appear to have lost SP isoforms as mentioned above, as all of the isoforms from these two species cluster on a single branch within the QAE grouping of the gene phylogeny. The serum of P. brachycephalum isolated from McMurdo Sound was found to contain only two QAE isoforms [41] whereas transcriptome sequencing (over 480k sequences) from fishes isolated over 4000 km away in the South Shetland Islands [50] revealed three QAE isoforms, two of which were unique, but no SP isoforms. The isoform encoded by the most abundant transcript matched the predominant isoform isolated from the McMurdo fishes. Interestingly, this isoform has two mutations that convert the ice-binding surface from QAE-like (TLV) to SP-like (TPA). Although it was reported to be fully active [41], this is no longer a certainty as we now know that trace amounts of fully-active QAE isoforms can confer full activity to SP isoforms [27, 78]. Why then would fish retain isoforms that are not fully active? One observation is that the SP-like mutations to the ice binding surface appear to extend the footprint of the pyramidal-plane binding surface while slightly reducing the overall hydrophobicity of the protein. This may strengthen or speed up ice binding while improving solubility of the AFP. Additionally, the AFPs are substantially more hydrophobic than Antarctic eelpout SAS-B, a difference that was also noted when the NMR structure of the corresponding domain of human SAS was solved [55]. While this invariably relates to function, as hydrophobicity is a general property of the IBS of type III AFPs and loss of the hydrophobic residues flanking the IBS can be deleterious [84], it may be that a mixture of QAE and SP isoforms is more soluble at high concentrations than one isoform alone. For example, a notched-fin eelpout SP isoform expressed in bacteria has shown a propensity to dimerize [53]. Furthermore, it seems unlikely that the ice-binding site of the dominant Antarctic QAE isoform would have mutated to an SP-like ice-binding site if there was no selective advantage to having both types.

The lack of an SP isoform in the Antarctic eelpout also seems probable, as sequencing of 61 AFP-encoding expressed sequence tags (ESTs) and six BAC clones containing multiple AFP genes [39, 40] failed to turn up a single SP isoform. The various isoforms from this species have a number of mutations on the ice-binding surface, such as TLI or TLM instead of TLV, but more importantly, this is the only species known to produce tandemers (sequential AFP domains encoded in a single transcript) [37, 39, 40]. This again suggests that fully-active QAE monomers alone might not be as effective as a mixture of isoforms and reveals yet another mechanism by which the diversity of the type III AFP family is attained.

It is tempting to try to deduce function from the sequences of the various isoforms of type III AFP, but unfortunately, the effect of the various mutations on AFP activity are very difficult to predict. Mutations of SP isoform notched-fin eelpout-S1, introduced to mimic the ice-binding residues of a fully-active QAE isoform (TPA to TVL), almost restored wild type activity. However, the crystals still grew, albeit at a much slower rate [78]. This may be because other residues are also know to affect activity. For example, a reduction in the size of hydrophobic residues bordering the ice-binding surface of a QAE isoform reduced activity [84]. Additionally, it is known that some QAE isoforms are not fully active [27, 28]. Therefore, it may be that the activities of closely-related sequences could vary and the only reliable way to determine this would be to express and purify these isoforms.

There are over 400 known species of Zoarcales [66], including others that inhabit the ocean near Antartica [67], so the sampling of 12 species catalogued in this study by us and others is by no means a comprehensive survey of all of the type III diversity that is present in this infraorder. There may well be northern species that lack SP isoforms or southern species that have retained them, although the two Antarctic species that have been characterized do appear to have lost their SP isoforms and duplicated/amplified their QAE isoforms. Lineage-specific amplifications of SP isoforms have also taken place in northern species. Additionally, we have shown that all of the known isoforms were derived from a single progenitor sequence and have further elucidated the variability of this gene family in this infraorder. The addition from this study of 24 sequences and their polymorphisms to the type III AFP family should also prove useful in modelling or mutagenesis studies aimed at further understanding the relationship between structure and function in type III AFPs, particularly as a number of these isoforms have mutations on or near the ice-binding surface that would be expected to alter their activity. In addition, we have shown that the type III AFP arose early in the Zoarcales lineage and this may have been one of the factors that allowed this group of fishes to migrate into waters across the globe and to diversify over the last ~20 Ma into the over 400 species that are extant today.

Materials and methods

Supporting information