Cold Adaptation of Microtubule Assembly and Dynamics

The microtubules of Antarctic fishes, unlike those of homeotherms, assemble at very low temperatures (−1.8 °C). The adaptations that enhance assembly of these microtubules are intrinsic to the tubulin dimer and reduce its critical concentration for polymerization at 0 °C to ∼0.9 mg/ml (Williams, R. C., Jr., Correia, J. J., and DeVries, A. L. (1985)Biochemistry 24, 2790–2798). Here we demonstrate that microtubules formed by pure brain tubulins of Antarctic fishes exhibit slow dynamics at both low (5 °C) and high (25 °C) temperatures; the rates of polymer growth and shortening and the frequencies of interconversion between these states are small relative to those observed for mammalian microtubules (37 °C). To investigate the contribution of tubulin primary sequence variation to the functional properties of the microtubules of Antarctic fishes, we have sequenced brain cDNAs that encode 9 α-tubulins and 4 β-tubulins from the yellowbelly rockcod Notothenia coriiceps and 4 α-tubulins and 2 β-tubulins from the ocellated icefish Chionodraco rastrospinosus. The tubulins of these fishes were found to contain small sets of unique or rare residue substitutions that mapped to the lateral, interprotofilament surfaces or to the interiors of the α- and β-polypeptides; longitudinal interaction surfaces are not altered in the fish tubulins. Four changes (A278T and S287T in α; S280G and A285S in β) were present in the S7-H9 interprotofilament “M” loops of some monomers and would be expected to increase the flexibility of these regions. A fifth lateral substitution specific to the α-chain (M302L or M302F) may increase the hydrophobicity of the interprotofilament interaction. Two hydrophobic substitutions (α:S187A in helix H5 and β:Y202F in sheet S6) may act to stabilize the monomers in conformations favorable to polymerization. We propose that cold adaptation of microtubule assembly in Antarctic fishes has occurred in part by evolutionary restructuring of the lateral surfaces and the cores of the tubulin monomers.

The capacity of the cytoplasmic tubulins of Antarctic fishes to form microtubules at temperatures as low as Ϫ1.8°C, the freezing point of the seawater habitat of these fish, is remarkable. The critical concentration for polymerization at 0°C of purified, MAP 1 -free brain tubulin is ϳ0.9 mg/ml (1)(2)(3), which approximates the values observed for mammalian brain tubulins at 37°C (1,4). Unlike mammalian microtubules, however, the cytoplasmic microtubules of Antarctic fishes are very stable and exhibit slow dynamics at cold temperatures (5,6). Evidently, both the capacity to polymerize at low critical concentration and the slow exchange of tubulin dimers into and out of microtubule polymers are properties intrinsic to the tubulin subunits of these fishes. These unusual functional characteristics must be explicable by structural alterations to the primary sequences and/or by changes in the posttranslational modifications of the tubulin chains that, alone or in concert, modify the quaternary interactions between dimers in a microtubule. Both the longitudinal, intraprotofilament and the lateral, interprotofilament contacts may be subject to structural remodeling, and adaptation might require coordinated changes located in the complementary contact domains of the tubulin chains.
Disentangling the contributions of isotype sequence variation and differential posttranslational modification to cold adaptation of microtubule assembly ideally requires comprehensive functional and structural analyses of tubulins that derive from comparable tissues of related cold-and warm-living taxa. Among vertebrates, the notothenioid fishes of the Southern Ocean and mammals experience body temperatures (Ϫ1.8 and ϩ37°C, respectively) that differ by almost 40°C, and their brains provide abundant quantities of tubulins composed of multiple ␣ and ␤ isotypes (7). By using representative species from these taxonomic groups as models, we compare in this report the dynamics of their microtubules at cold and warm temperatures and determine via cDNA cloning the primary sequence variations that distinguish their tubulin isotypes. By mapping the residue substitutions found commonly in the Antarctic fish isotypes, but rarely in other vertebrates, onto the three-dimensional structures of the tubulin dimer and the microtubule (8,9), we deduce potential structural interactions that predispose the fish tubulins to form stable, relatively non-dynamic polymers at cold temperatures. Our analysis of the role of posttranslational modification in cold adaptation of microtubule assembly will be presented elsewhere. 2

EXPERIMENTAL PROCEDURES
Collection of Fish Tissues-Specimens of the yellowbelly rockcod, Notothenia coriiceps, of the humped rockcod, Gobionotothen gibberifrons, and of the ocellated icefish, Chionodraco rastrospinosus, were collected by bottom trawling from the R/V Hero or from the R/V Polar Duke near Low and Brabant Islands in the Palmer Archipelago. They were transported alive to Palmer Station, Antarctica, where they were maintained in seawater aquaria at Ϫ1.5 to ϩ1°C. Brain tissues were dissected and used immediately for preparation of microtubules or RNA. Some brains were frozen in liquid nitrogen and maintained at Ϫ70°C for later processing (RNA only).
Preparation of Tubulins-Tubulin from fresh brain tissues of G. gibberifrons or N. coriiceps was purified by DEAE ion-exchange chromatography and one cycle of microtubule assembly as described by Detrich et al. (2,10). Bovine brain tubulin was prepared as described by Gildersleeve et al. (11).
Video Microscopy of Microtubule Dynamics-Video-enhanced differential interference-contrast light microscopy was carried out by modifications of techniques described by Gildersleeve et al. (11). Samples were observed in closed chambers of approximately 20-m thickness. Tubulin, sea urchin flagellar axonemal seeds, and buffer were mixed at 0°C, a sample was applied to a clean slide and covered with a coverslip, and the chamber was sealed with Valap. The slide was then mounted on the stage of the microscope for observation. Tubulin concentrations (0.1-1.0 mg/ml for fish brain, 1.4 mg/ml for bovine brain) were chosen at which axoneme-mediated nucleation of microtubule assembly occurred efficiently, whereas self-nucleation of free microtubules was suppressed. For measurements at 5°C, the microscope was placed in a temperature-controlled enclosure located in a cold room (1-2°C); a thermostated heater and fan maintained the temperature within the enclosure at 5 Ϯ 1°C. For observations at 25 or 37°C, the microscope and its enclosure were placed at room temperature (20°C), and the heater thermostat was adjusted to maintain the desired supra-ambient temperature. Videotapes of microtubules were recorded, and rates of microtubule growth and shortening were analyzed by fitting, according to the method of least squares, straight lines to moving windows of data points (covering length changes of approximately 0.5 mm and intervals of several minutes) as described by Gildersleeve et al. (11).
Protein Determinations-Tubulin concentrations were measured by the method of Bradford (12) with bovine serum albumin as the standard.
cDNA Library Construction and Screening-The production of two N. coriiceps brain cDNA libraries, the first primed with oligo(dT) and constructed in gt10 and the second primed with random hexanucleotides/oligo(dT) (75:25%) and cloned in ZAP II (Stratagene), has been described previously (13,14). The C. rastrospinosus brain cDNA library was constructed in gt10 by similar methods, with cDNA synthesis primed by a 50:50 mixture of random hexanucleotides and oligo(dT).
Libraries were screened for recombinant clones containing ␣and ␤-tubulin coding sequences as described by Parker and Detrich (14). One hundred fifty nine candidate ␣-tubulin cDNA isolates and 55 prospective ␤-tubulin cDNAs were obtained from a total of six screens of the two N. coriiceps libraries (632,000 total recombinant phage used in three ␣-screens, 385,000 for three ␤-screens). Thirty primary isolates for ␣-tubulin cDNAs, and 30 for ␤, were picked from single screens of the C. rastrospinosus library (240,000 phage examined for ␣ and 12,000 for ␤). After tertiary plaque purification and screening, clones containing the longest inserts (Ն1 kilobase pair) were selected for subcloning and sequence analysis.
Subcloning and DNA Sequence Analysis-cDNA inserts from N. coriiceps tertiary phage clones were subcloned either by restriction digestion and ligation into pUC19 (gt10 library) or by in vivo excision into pBluescript (ZAP II). Subcloned cDNAs were sequenced on both strands by primer walking using either manual or automatic methods. Manual sequencing was performed using the dideoxynucleotide chain termination method (15), T4 DNA polymerase (Sequenase II; United States Biochemical Corp.), and conventional polyacrylamide gels. Automatic sequencing employed the PRISM Ready Reaction Dye Deoxy Termination Cycle Sequencing Kit (Applied Biosystems) and an Applied Biosystems 373A automated DNA sequencer (University of Maine DNA Sequencing Facility). Tertiary phage clones containing C. rastrospinosus cDNA inserts were sequenced directly using the automated sequencer.
The nucleotide sequences of the ␣and ␤-tubulin cDNAs and the primary sequences of the encoded proteins were analyzed by use of the Clustal method provided by DNASTAR MegAlign. Comparisons of the sequences of the Antarctic fish tubulins to those of ␣and ␤-tubulins from other organisms (GenBank TM and Swiss-Prot accession numbers accompany text) were performed using the BLASTP program (National Center for Biotechnology Information).
Three-dimensional Analysis of Tubulin Sequence Substitutions-The polarity difference map of the lateral contact interfaces of Antarctic fish and bovine tubulins was generated using GRASP (19). The ribbon diagram of interprotofilament interactions between Antarctic fish tubulins was produced with MOLSCRIPT (20). Fig. 1 presents temporal profiles of the length changes for typical individual microtubules composed of pure tubulins from Antarctic fish brain at 5 (A) and 25°C (B) and from bovine brain at 37°C (C). Fig. 2 shows histograms, gathered by measuring many such microtubules, of the observed rates of growth and shortening of fish brain (A and B) and bovine brain (C) microtubules under these conditions. Table I summarizes the mean rates of growth and shortening, and the frequencies of transition between the two phases, which together govern the dynamics of the polymers.

Dynamics of Microtubules from Antarctic Fishes at Near-and Supra-physiological Temperatures-
From these data, it is apparent that the MAP-free brain microtubules of two Antarctic fishes are substantially less dy-namic than MAP-free bovine microtubules. First, the rates of growth and shortening of fish brain microtubules at the nearphysiological temperature of 5°C were 1-2 orders of magnitude smaller than those observed for brain microtubules of the cow at its body temperature, 37°C. (The dynamics of the microtubules from the two fish species were indistinguishable.) Second, the frequencies of catastrophe (transition from growth to shortening) and rescue (transition from shortening to growth) of the fish microtubules were more than an order of magnitude smaller than those observed for the bovine structures. Third, the rates of growth and shortening of the fish microtubules remained small even at the supra-physiological temperature of 25°C. 3 Together, these results indicate that the sluggish dynamics and pronounced cold stability of brain microtubules from Antarctic fishes must derive from structural features intrinsic to their tubulin subunits.
Primary Sequences of Brain ␣and ␤-Tubulins from Two Antarctic Fishes-To evaluate whether primary sequence variation contributes to cold adaptation of microtubule assembly, we have cloned cDNAs that encode most of the brain tubulin chains of the Antarctic nototheniid N. coriiceps. To date, nine distinct cDNAs encoding ␣-tubulins (NcTb␣1-NcTb␣9) and four encoding ␤ isotypes (NcTb␤1-NcTb␤4) have been obtained. We have also sequenced a subset of tubulin cDNAs from the channichthyid C. rastrospinosus (four ␣ and two ␤) to determine if a common pattern of sequence changes characterizes the tubulins of fish species representing two different notothenioid families. Fig. 3 shows the ␣and ␤-tubulin sequences of the two fishes, compared with consensus sequences for vertebrate ␣and ␤-tubulin chains and positioned with respect to the secondary structural elements of the tubulin monomer (8). Residue substitutions that distinguished the Antarctic fish chains from most of the tubulins of other vertebrates are coded by color to indicate their effects on local polypeptide hydrophobicity (green, increased; blue, decreased; yellow, no change). Several of the C. rastrospinosus ␣and ␤-tubulins corresponded to isotypes from N. coriiceps and are numbered to indicate this identity (e.g. CrTb␣7 and -␣9 are identical to NcTb␣7 and -␣9). The discovery of C. rastrospinosus tubulin cDNAs (CrTb␣10 and -␣11, CrTb␤5) that have not been found in N. coriiceps suggests that our screens of the latter fish were not exhaustive. Alternatively, some tubulin genes may have diverged slightly subsequent to the evolutionary separation of the rockcods and icefishes approximately 7-15 million years ago (21). The three panels present selected measurements showing typical behavior for Antarctic fish brain microtubules at 5 (A) and at 25°C (B) and for bovine brain microtubules at 37°C (C). For clarity, initial microtubule measurements have been offset arbitrarily on the y axis. Note that for both kinds of microtubules, the rates of growth and shortening are quite variable, a phenomenon long recognized in mammalian microtubules (11). Note also the small differences between the three panels in the horizontal and vertical scales. In each case, MAP-free tubulin was assembled, with the use of sea urchin axonemes as nuclei, in PMD buffer ϩ 1 mM GTP (see "Experimental Procedures"). Observations of individual microtubules were continued for 20 -60 min.
FIG. 2. Frequency histograms of rates of microtubule growth and shortening. The three panels show distributions of observed rates of growth and shortening for Antarctic fish brain microtubules at 5 (A) and at 25°C (B) and for bovine brain microtubules at 37°C (C; inset, growth rates on an expanded x axis). The broad distribution of rates in each of the three cases corresponds to the visible roughness of the observed length-versus-time curves in Fig. 1 (11). Note the large difference in the scales of the x axes.
␤-Tubulin isotypes are strongly conserved among advanced vertebrate taxa (birds and mammals), with distinct classes largely defined by shared, carboxyl-terminal signature sequences (residues 431 to end) (7,22,23). (␣-Tubulin chains, by contrast, tend to show greater interspecific variability.) Is the strict conservation of ␤-chains also characteristic of fishes, the most phylogenetically primitive of vertebrates? Previously, we identified NcTb␤1 as a possible class II isotype but noted that assignment as a class IVb type was nearly as plausible (13). With a nearly complete set of ␤-tubulin cDNAs from two Antarctic fishes in hand, we can now address the question more definitively. NcTb␤4 was classified as a neuron-specific ␤ III isotype based on its long carboxyl terminus that ends in a basic motif (VRHDVRH) and on several conserved internal substitutions (Ser 35 and Tyr 436 ). With the ambiguous exception of NcTb␤1, the remaining ␤-tubulins (NcTb␤2 and -␤3, CrTb␤2 and -␤5), based on their carboxyl-terminal sequences, were most closely related to the vertebrate class IVb isotype (Table  II); consideration of conserved, coordinated substitutions at isotypic "hot spots" (7) failed to differentiate these chains into other ␤-classes. Furthermore, each of the ␤-tubulins, excluding NcTb␤4, contained the axonemal signal sequence EGEFXXX (positions 433-439 of vertebrate class IVb; X ϭ acidic residue). Thus, the interspecific conservation of ␤-chain isotypes that typifies higher vertebrates appears not to hold for the more distantly related fishes. Three brain ␤-tubulin sequences (␤1-␤3; GenBank TM accession numbers AF102890, AAD56401, and AF184595, respectively) from the Atlantic cod (Gadus morhua), a northern marine fish that typically lives at 8 -15°C but is able to acclimatize to near-freezing temperatures in winter, have been described (24). Although the southern notothenioids and northern gadoids diverged long before their respective oceans cooled (25,26), some of the unique ␤-chain substitutions of the Antarctic fishes were present in one or more of the gadid chains (e.g. ␤:Y202F of cod ␤2; Table III). The significance of the conserved notothenioid/gadoid changes will be considered below (see "Mapping Residue Substitutions to the Structure of the Tubulin Dimer" and "Discussion").
Steady-state Expression of Brain ␣and ␤-Tubulin Genes-The neural tissues of vertebrates typically express four or more ␣-tubulin genes and a comparable number of ␤ (7,22,23). Our results indicate that Antarctic fishes express 10 -11 ␣and 4 -5 ␤-tubulin genes in their brain tissues. To estimate the relative, steady-state abundances of the corresponding tubulin isotypes, we hybridized Northern slot blots of total brain RNA from N. coriiceps to probes specific to the 3Ј-UTRs of 7 ␣and 4 ␤-tubulin cDNAs. 4 Fig. 4 shows that NcTb␣2, -␣7, -␣8, and -␣9 mRNAs were prominent components of the ␣-tubulin transcript pool, whereas NcTb␣3, -␣5, and -␣6 were considerably less prevalent. Among ␤-chain transcripts, mRNAs for NcTb␤2 and -␤3 were abundant and roughly equivalent in amount, the NcTb␤1 message was less prominent, and the NcTb␤4 transcript, corresponding to the ␤ III isotype, constituted a minor component. Under the assumption that steady-state transcript levels mirror the relative quantities of the polypeptides generated by translation and disregarding the contribution of posttranslational modification, we suggest that the brain tubulins isolated from Antarctic fishes contain four major ␣and two major ␤-tubulin chains. It is noteworthy that the more abundant tubulin chains almost universally contained the novel sequence substitutions shown in Fig. 3.
Mapping Residue Substitutions to the Structure of the Tubulin Dimer-The solution of the tertiary structure of the tubulin dimer (8) and of the quaternary interactions between dimers within a microtubule (9) provides the framework for interpretation of the unusual primary sequence variations of the brain tubulins of Antarctic fishes. We consider these changes first in the context of potential remodeling of the longitudinal and lateral contact surfaces of the dimer and second with respect to their alteration of the internal structure of the tubulin monomers.
To our surprise and contrary to our prior predictions (13,27), none of the residue changes that characterize the Antarctic fish tubulins mapped to the longitudinal, interdimer interfaces. Rather, surface substitutions were restricted to loops of the lateral, interprotofilament contacts and the luminal sides of the monomers. Fig. 5 presents a polarity difference map of the lateral contacts, with residue changes coded by color as before (green, increased hydrophobicity; blue, decreased; yellow, no change). Four of the five substitutions (␣-chain: A278T and S287T; ␤-chain: S280G and A285S) at the lateral contact faces were found in the S7-H9 "M" (microtubule) loops of some, but not all, of the ␣ and ␤ isotypes (cf. Fig. 3). Relative to their structure-forming tendencies (28), most of these changes would be expected to introduce greater flexibility into the "hinge points" of the M loops. Because the M loop of the ␤-chain is thought to interact with the nucleotide-sensitive H3 helix of the neighboring ␤-subunit to control the strength of the lateral interaction between dimers (9), increased flexibility in this region may strengthen interprotofilament contacts and slow the conversion of the "straight" dimers of a growing microtubule end to the "curved" dimers of the shortening state (29,30).
Previously, we suggested that evolution of their tubulins to form greater numbers of hydrophobic interactions at sites of interdimer contact may cause the greater entropic control of microtubule assembly in Antarctic fishes (2, 3). The fifth, ␣-specific lateral substitution, M302L,F, was distinctly hydrophobic in character (cf. Ref. 31) and was present in most of the fish subunits (Fig. 3). Found adjacent to Val 303 (and near Val 204 ) at the base of a hydrophobic pocket formed by the H9 -S8 loop, M302L,F may contribute to immobilization of additional water molecules by the dimer. Displacement of this water upon lateral association provides a plausible rationale for the unexpected energetics (1-3) of polymerization of Antarctic fish tubulins. If operative, the mode of displacement (perhaps local conformational changes in the ␣-chain induced by contact with the H3 helix of the ␣ monomer in the adjacent protofilament) remains uncertain. Fig. 6 shows a ribbon diagram of two dimers interacting laterally. Highlighted are both the residue changes of the lateral contact loops and substitutions within buried elements of secondary structure. 5 The latter changes, some of which occur consistently in many of the fish isotypes (␣:S187A in helix H5, ␤:Y202F in sheet S6; see Fig. 3), would be expected to increase the hydrophobicities of the monomeric cores. We postulate that ␤:Y202F may be particularly important for strengthening the lateral contacts between microtubule protofilaments. This residue, which is shared by one of three Atlantic cod ␤ isotypes (␤2; GenBank TM accession number AAD56401), resides near the interface between the amino-terminal GTP-binding domain 5 Note that side chains shown belong to the vertebrate consensus but have been coded per Fig. 3 to indicate the polarity change introduced by the fish substitutions.

FIG. 3-continued
␤:Y202F may be acting as a surrogate taxol to stabilize a GTP-like structure of tubulin that results in strong lateral interactions, even at low temperatures. DISCUSSION Microtubules assembled from pure brain tubulins of coldliving fishes, unlike those of homeotherms, are stable at temperatures between 0 and 25°C and exhibit limited dynamic instability (see Refs. 5 and 6 and this report). Slow dynamics must therefore result from structural changes intrinsic to the tubulins of these organisms that modify the quaternary inter-actions between dimers in a microtubule. To investigate the contribution of tubulin primary sequence variation to the unique functional properties of the microtubules of Antarctic fishes, we have sequenced a total of 11 ␣and 5 ␤-tubulin cDNAs from two representative species of Antarctic notothenioid fishes. We find that these tubulins contain small sets of unusual residue substitutions that map to the lateral, interprotofilament surfaces or to the interiors of the ␣and ␤ polypeptides. By contrast, the longitudinal, intraprotofilament interaction surfaces appear not to be altered.
Evolution of Tubulin Isotypes to Enable Microtubule Assembly at Low Temperature: All Together Now?-The Southern Ocean began to cool approximately 25-40 million years ago and reached its current freezing temperature during the midlate Miocene (5-14 million years ago (34,35)). Subjected to the selective pressure of an increasingly cold thermal environment, the coastal fishes of the Antarctic diverged from temperate fishes (36) and evolved molecular, cellular, and physiological adaptations that maintain metabolic efficiency and preserve macromolecular function in their now chronically cold marine habitat (Ϫ1.8 to ϩ1°C). One may ask whether the primary sequences of each of the many brain ␣and/or ␤-tubulin isotypes of Antarctic fishes need be altered equivalently to ensure that microtubules polymerize efficiently at cold temperatures. Probably not. Wallin and Billger (37) have shown that brain TABLE II Isotypic classification of ␤ tubulins from Antarctic fishes The carboxy-terminal sequences of ␤-tubulins from N. coriiceps and C. rastrospinosus (residues 431 to carboxyl terminus) were compared with the corresponding isotype-defining sequences of neural ␤-chains (classes I-IV) from higher vertebrates. For each pairwise comparison, sequence similarity was calculated as percentage residue identity with respect to the longer sequence. Asterisks indicate gaps introduced to maximize sequence similarity. The longer class III carboxyl termini (c␤4, Nc␤4) are shown unaligned and in reduced point size.

EEEENFDEEAD-EEIA
a Question marks indicate residues that are unknown because the CrTb␤5 cDNA lacked some 5Ј-coding sequence.
FIG. 4. Expression of brain tubulin isotypes in N. coriiceps. Steady-state transcript levels of individual N. coriiceps ␣and ␤-tubulin isotypes were evaluated by hybridization of total brain RNA (5 g/slot applied to nylon membranes by vacuum aspiration) to gene-specific probes derived from the 3Ј-UTRs of seven ␣-tubulin cDNAs (NcTb␣2, NcTb␣3, and NcTb␣5-NcTb␣9) and of four ␤-tubulin cDNAs (NcTb␤1-NcTb␤4) (see "Experimental Procedures"). tubulin or microtubule proteins from the cold-living Atlantic cod can recruit bovine brain tubulin dimers into hybrid microtubules in vitro at concentrations and temperatures that are non-permissive for assembly of the mammalian tubulin alone. Furthermore, two of three Atlantic cod ␤-tubulins (␤1 with the M loop substitution A285S and ␤2 with the Y202F substitution), when expressed in human HepG2 cells, confer partial cold tolerance on hybrid fish/human microtubules in vivo (38). Fields and Somero (39) have shown that cold adaptation of lactate dehydrogenase A 4 orthologs from Antarctic notothenioids can occur via different, but non-random, substitutions that increase the flexibility of regions adjacent to the active site. Our work suggests that evolutionary adaptation of brain tubulin ensembles for polymerization at psychrophilic temperatures results both from residue substitutions that are common to most ␣or ␤-chains and from other changes that are restricted to one or a few isotypes.
Qualitative or quantitative changes in the isotypic compositions of brain tubulins may also contribute to functional adaptation of microtubule assembly. We have shown here (Fig. 4) and elsewhere (3,6,40) that mRNAs and proteins corresponding to the ␤ III isotype are minor species in Antarctic fish brain. Bovine brain tubulin depleted of the ␤ III isotype by immunoaffinity chromatography polymerizes more readily than does unfractionated tubulin (41), consistent with an inhibitory effect of ␣␤ III dimers on assembly of the brain tubulin ensemble. Furthermore, microtubules composed of ␣␤ III dimers are considerably more dynamic than microtubules containing other specific ␤ isotypes (42). Hence, the paucity of the ␤ III isotype in the brain tubulins of Antarctic fishes may contribute to the coldassembling and slowly dynamic phenotype of their microtubule polymers.
Cold Adaptation of Microtubule Assembly Via Small Numbers of Primary Sequence Changes to the ␣and ␤-Tubulins: Potential Mechanisms-We envision that the residue substitutions present in the ␣and ␤-tubulin subunits of Antarctic fishes predispose the dimer to form microtubules by three potentially synergistic mechanisms. First, the lateral interactions between adjacent protofilaments may be strengthened by the flexibility introduced into the ␣and ␤-tubulin M loops. Nogales et al. (9) have proposed that the amino-and carboxylterminal parts of the M loops function as hinges that can accommodate variation in microtubule protofilament number. Although not universally present in the ␣ and ␤ monomers of the Antarctic fishes, four of the five substitutions that we have detected at the lateral interfaces of the tubulin dimer (two in ␣ and two in ␤) occur at or near the potential M loop hinge points. Relaxation of the M loop conformation, particularly in the ␤-subunit, may enable the ␤/␤ lateral interaction to accommodate to the probable hydrolysis-dependent conformational change in H3, thus mitigating its destabilizing effect. It is interesting that many cold-sensitive, charged-to-alanine mutants of yeast ␣and ␤-tubulins map to their respective M loops, H3 helices, and H1-S2 loops, the major contributors to lateral, interprotofilament contact (43,44). Second, provision of additional hydrophobic interactions at the lateral surface of the ␣-chain would strengthen interprotofilament interaction at low temperature and explain the greater entropic control of the polymerization of the fish tubulins. The near universal substitution of Leu or Phe for Met at position 302 of the Antarctic fish ␣-chains may create a pocket of greater hydrophobicity that constrains additional water molecules whose release on subunit association would increase the entropy of polymerization. Similarly, ␣-tubulins from two psychrophilic algae of the genus Chloromonas contain lateral hydrophobic substitutions (M268V and A295V relative to the ␣-tubulins of the temperate alga Chlamydomonas reinhardtii) that are likely to increase FIG. 6. Ribbon diagram of tubulin dimers in adjacent microtubule protofilaments. Mapped onto the structure of the bovine tubulin dimer are external and internal residue substitutions found commonly in the tubulins of Antarctic fishes but rarely in the isotypes of other vertebrates. The side chains shown, which correspond to those of the vertebrate consensus residues, are color-coded to indicate the polarity change introduced by the fish substitution: green, increased hydrophobicity; blue, decreased hydrophobicity; yellow, minimal polarity change. Substitution nomenclature: Vertebrate consensus residue/sequence position/Rockcod residue. The GXP nucleotides are shown in orange. Rockcod amino acid substitutions that increase or decrease surface hydrophobicity with respect to bovine tubulin are indicated in green and blue, respectively, whereas those that cause minimal polarity change are shown in yellow. Substitution nomenclature: vertebrate consensus residue/sequence position/Rockcod residue. The vertebrate consensus sequences serve as proxies for those of bovine tubulin (cf. Nogales et al. (8)). the hydrophobicity of their lateral contact surfaces (45). Third, the ␤:Y202F substitution may stabilize the assembly-competent conformation of tubulin. Assuming that structural rearrangement between the amino-terminal and central domains of the ␤ monomer converts the straight, protofilament-forming conformation of the dimer to the curved, depolymerizing state, then the Phe 202 substitution at the interface between the two domains may act to resist the conformational change normally induced by nucleotide hydrolysis and phosphate release. Thus, ␤:Y202F may function like taxol to stabilize a GTP-like structure that forms strong lateral interactions. Indeed, yeast tubulin, which also contains Phe 202 , has been observed to polymerize in vitro at temperatures as low as 4°C. 6 Definitive evaluation of our hypotheses will require the exploitation of a genetically tractable organism in which Antarctic fish tubulins, and site-directed mutants thereof, can be expressed abundantly for in vitro or in vivo analysis of the temperature dependence of microtubule assembly.
Convergent Evolution of Tubulin Sequence Alterations in Distantly Related, Cold-living Fish-The southern notothenioids and northern gadoids diverged well before the establishment of glacial conditions in their respective polar oceans (25,26), and their cold-adapted phenotypes evolved independently during different geologic epochs (46 -49). Nevertheless, the Antarctic rockcod N. coriiceps and the temperate/cold-living Atlantic cod G. morhua have evolved ␤-tubulins that share some M loop and internal residue substitutions (Table III), and their microtubules demonstrate similar dynamic properties (6). We propose that the two taxa have converged independently on a comparable evolutionary strategy to preserve microtubule assembly and function at cold temperatures. However, the eurythermal Atlantic cod, unlike the stenothermal notothenioids of the Southern Ocean, must acclimatize to seasonal variations in habitat temperature. One likely adjustment to maintain seasonal homeostasis of microtubule function would be the differential synthesis of ␤-chain isotypes containing Ser 285 and Ala 285 (e.g. cod ␤1 versus ␤3) and/or Phe 202 and Tyr 202 (e.g. cod ␤2 versus ␤3) under winter and summer conditions, respectively.
␤-Tubulin Isotypes and Microtubule Dynamics-Microtubules polymerized from the tubulins of chicken erythrocytes or of the budding yeast Saccharomyces cerevisiae, like those formed by Antarctic fish brain tubulins, demonstrate slow dynamics in vitro (50 -52). Yeast microtubules, for example, display rates of growth and shortening that are approximately a tenth those shown by bovine tubulin (51,52). Dynamics can be further suppressed in yeast tubulins by a mutation (T107K) introduced into the ␤-tubulin chain (52). Although the ␤-tubulins from Antarctic fish brain, chicken erythrocytes, and yeast cells are quite divergent, their primary sequences share a phenylalanine residue at position 202 in an otherwise conserved context (DETFCIDN). Thus, we propose that the repressed dynamics of the microtubules formed by these tubulins may be caused by a shared ␤-chain substitution whose effect is to strengthen lateral interactions by resisting conversion of the tubulin dimer to the curved, depolymerizing conformation. The potential importance of lateral interactions in governing dynamic instability has been emphasized in the "lateral cap" scheme (53). In that model, which successfully mimics many of the aspects of microtubule dynamics, it is changes in lateral affinity, more than changes in longitudinal interactions, which cause the microtubule to switch between growing and shortening phases.
Multifunctionality of Antarctic Fish ␤-Tubulins-The similarity of most Antarctic fish ␤-tubulins (excluding NcTb␤4) to the vertebrate class IVb ␤-chain (Table II) suggests that ␤ IVb may represent an ancestral, multifunctional isotype. This hypothesis is supported further by the presence of the "axonemal" signal sequence EGEFXXX (X ϭ acidic residue (7)) in each of the non-class III ␤-chains. Perhaps ␤-tubulin isotypes, with the exception of class III, had not evolved specialized functions prior to separation of the chordates from other metazoans. Fishes may subsequently have retained multifunctionality of their ␤-tubulins, whereas more advanced vertebrates and advanced invertebrate taxa evolved ␤ isotypes with distinct cellular functions (reviewed by Ludueñ a (7)).