The Interplay Among Cardiac Ultrastructure, Metabolism and The Expression of Oxygen-Binding Proteins in Antarctic Fishes

Antarctic icefishes (Channicthyidae) are one of six families within the suborder Nototheniodei that dominates both species number and biomass of fishes in the Southern Ocean (Eastman, 1993). Icefishes are unique among all vertebrates because as adults they lack the oxygen-binding protein hemoglobin (Hb). Because these fishes lack Hb, the oxygen-carrying capacity of their blood is only one-tenth of that of red-blooded teleosts (Ruud, 1954).

Channichthyids possess many unusual cardiovascular features that appear to compensate for the loss of circulating Hb. Their large heart-to-body mass ratio contributes to a mass-specific cardiac output that is 4–5 times greater than that of red-blooded teleosts (Hemmingsen et al., 1972). Blood volumes in icefish are 2–4 times greater than those of red-blooded teleosts, and they possess unusually large-diameter capillaries that minimize the peripheral resistance against which the heart must work (Hemmingsen and Douglas, 1970; Fitch et al., 1984). In combination, these cardiovascular characteristics provide a large blood volume that is circulated through the body at high flow to maintain oxygen delivery to working muscles.

The consensus has been that hearts from icefish also lack myoglobin (Mb), the oxygen storage and transport protein found in oxidative muscle (Hamoir and Geradin-Otthiers, 1980; Wittenberg and Wittenberg, 1989). Recent findings, however, have revealed the presence of this protein in heart ventricles of several species of icefishes (Sidell et al., 1997; Moylan and Sidell, 2000). Our laboratory has recently examined hearts from 13 of the 15 known species of channicthyid icefishes, and we have determined that myoglobin is expressed in eight of these species (Moylan and Sidell, 2000). During the evolution of the icefish family, expression of myoglobin has been lost through at least four independent mutational events, based upon patterns of myoglobin protein expression and phylogeny (Moylan and Sidell, 2000). The widely dispersed pattern of presence and absence of myoglobin within the channicthyid family initially suggested that the protein might not be functional at their cold body temperatures. Several recent studies, however, indicate that Mb is indeed functional in these fishes.

Kinetic analyses reveal that Mbs from icefish and other teleosts display faster rates of oxygen binding and dissociation at cold temperature than mammalian Mbs (Cashon et al., 1997). Experiments with isolated, perfused hearts from icefishes demonstrate that selective poisoning of Mb results in loss of mechanical performance by hearts that express the protein, but not in hearts that lack Mb (Acierno et al., 1997). These perfused heart experiments also show that hearts from species that naturally lack Mb are capable of meeting greater pressure/work challenges than hearts from icefish that express Mb in which the protein has been poisoned. These results strongly indicate that Mb is functional when present and that, to maintain cardiac function, the ultrastructural and/or metabolic characteristics of hearts lacking Mb have been modified to compensate for loss of the protein.

Johnston and Harrison (1987) compared the ultrastructure of the heart ventricle between a myoglobinless icefish, Chaenocephalus aceratus, and a red-blooded nototheniid, Notothenia neglecta. They determined that hearts from Chaenocephalus aceratus had significantly higher mitochondrial densities than those from N. neglecta and hypothesized that these high densities might enhance intracellular oxygen diffusion in hearts of species lacking Mb. Chaenocephalus aceratus and N. neglecta, however, differ in their expression of both oxygen-binding proteins, Hb and Mb. Thus, whether architectural differences observed between these hearts are correlated with the loss of Mb or Hb expression was not definitively resolved.

A recent description of the pattern of both Mb and Hb expression among Antarctic notothenioid fishes now permits us to differentiate between structural and metabolic characteristics that are specifically correlated with the loss of Mb and those correlated with the loss of Hb. We examined heart ventricle in two species of channicthyid icefishes, Chaenocephalus aceratus (−Hb/−Mb) and Chionodraco rastrospinosus (−Hb/+Mb), and a closely related red-blooded nototheniid Gobionotothen gibberifrons (+Hb/+Mb). The ultrastructure of cardiac muscles was examined using electron microscopy, and cellular structures were quantified using stereological techniques. We also measured the maximal activities of key enzymes from several metabolic pathways as indices of the metabolic capacities of the tissues. Because all three species are phylogenetically closely related and ecotypically similarly sluggish, demersal fishes, we are confident that differences observed in cardiac muscle can be attributed to differences in the expression of oxygen-binding proteins rather than to lifestyle or genetic distance.

Gobionotothen gibberifrons, Chionodraco rastrospinosus and Chaenocephalus aceratus were captured using an otter trawl deployed from the R/V Polar Duke in Dallmann Bay (64°N, 62°W) at approximately 150 m depth during the austral autumn of 1991, 1993, 1995 and 1997 and the winter of 1996. Animals were maintained in shipboard circulating seawater tanks and transported to the US Antarctic Research Station, Palmer Station, on Anvers Island. Here, they were transferred to the Palmer Station aquarium and maintained unfed in covered and circulating seawater tanks at 0±0.5 °C.

Fishes were killed by a sharp blow to the head. The hearts were quickly excised and placed in an ice-cold solution (260 mmol l⁻¹ NaCl, 2.5 mmol l⁻¹ MgCl₂, 5.0 mmol l⁻¹ KCl, 2.5 mmol l⁻¹ NaHCO₃, 5.0 mmol l⁻¹ NaH₂PO₄, pH 8.0) and allowed to contract for several minutes to clear them of blood. They were then placed in an ice-cold fixative solution (3 % glutaraldehyde, 0.1 mol l⁻¹ sodium cacodylate, 0.11 mol l⁻¹ sucrose and 2 mmol l⁻¹ CaCl₂, pH 7.4) and perfused with fixative retrogradely through the bulbous arteriosus using a peristaltic pump. The pump was fitted with small-diameter rubber tubing and secured within the bulbous arteriosus using surgical silk. Hearts were perfused for 1 min at a flow rate of 15 ml min⁻¹ and then for 30 min at a flow rate of 9 ml min⁻¹. They were then stored in fixative at 4 °C for 8–10 h, with a change of fixative after the initial 4–6 h. Hearts were then transferred into Trumps buffer (1 % glutaraldehyde, 4 % formaldehyde, 0.1 mol l⁻¹ sodium cacodylate, 0.11 mol l⁻¹ sucrose and 2 mmol l⁻¹ CaCl₂, pH 7.4) and stored at 4 °C until they were transported to our laboratory at the University of Maine.

Ventricles were cut in half lengthwise, and a transmural section spanning from epicardium to endocardium was excised from the center of one half of each heart. Each transmural section was then subdivided into three regions: blocks closest to the epicardium (V1), blocks within the myocardium (V2) and blocks nearest the endocardium (V3). Blocks were rinsed briefly in an ice-cold solution (0.1 mol l⁻¹ sodium cacodylate, 7 % sucrose, 2 mmol l⁻¹ CaCl₂, pH 7.4), and then rinsed for 30 min and stored overnight at 4 °C in the solution. Blocks were post-fixed in an ice-cold solution of 1 % osmium tetroxide, 0.1 mol l⁻¹ sodium cacodylate, 7 % sucrose and 2 mmol l⁻¹ CaCl₂, pH 7.4, for 1.5 h, briefly rinsed in reagent-grade water, dehydrated through a series of increasing concentrations of ethanol (70 %, 95 %, 100 %) and cleared with propylene oxide. Blocks were stored overnight at room temperature in a mixture of propylene oxide:resin (2:1) with the lids slightly ajar to allow the propylene oxide to evaporate slowly. Blocks were then infiltrated with a mixture of Epon and Araldite resin for 1 h under vacuum, with a change of resin after the initial 30 min, and cured at 60 °C for 48 h. Hearts were fixed on site at Palmer Station during March and April 1995 and post-fixed at the University of Maine between June and July 1995.

Initially, blocks from each of the three regions of ventricle described above were sampled in two animals from each of the three species. Ultrastructural variables were quantified to determine the amount of variation in these variables within the ventricle. Because no variation was detected, an additional four blocks, one per individual, were randomly chosen from each of the three species so that a total of 10 blocks per species were analyzed from six individuals.

Blocks were first thick-sectioned (1.5 μm) with an LB4 microtome to verify the integrity of the tissue. Sections were stained with 1 % Toluidine Blue in 1 % sodium borate for 30 s on a warm plate. Blocks were then trimmed and thin-sectioned using a diamond knife and Sorvall MT2-B ultramicrotome. Sections were collected on 400 mesh copper grids and stained with 2 % uranyl acetate followed by 0.5 % lead citrate.

Sections were viewed with a Philips CM-10 transmission electron microscope equipped with a tilting goniometer stage. The stage was adjusted to 0 ° each time, ensuring that the beam was consistently perpendicular to the grid. Ten micrographs were taken at a magnification of 5200× for quantifying mitochondrial surface and volume densities and myofibril volume densities. Ten to twelve micrographs were taken at a magnification of 39 000× for measuring mitochondrial cristae surface densities. Micrographs were taken using the aligned systematic quadrats subsampling method (Cruz-Orive and Weibel, 1981). Individual mitochondria with the most clearly defined inner mitochondrial membrane were chosen for micrographs from within each randomly chosen field of view at a magnification of 39 000×. Calibration grids were photographed at each magnification to calculate final magnifications.

Mitochondrial and myofibril volume densities were quantified using point-counting methods; mitochondrial surface densities were measured using the line-intercept technique (Weibel, 1979). Micrographs were projected onto a Summagraphics II digitizing tablet at a final magnification of 13 400×. Images were overlaid with a square lattice test pattern with spacing equal to 1.34 μm on projected micrographs. Care was taken to exclude epithelial cells, endothelial cells, blood cells, extracellular matrix and luminal spaces from the measurements.

Mitochondrial cristae surface densities were quantified at a final magnification of 96 000× using the line-intercept method (Weibel, 1979). Micrographs of G. gibberifrons mitochondria were printed for best resolution of mitochondrial inner membrane. Regions of the mitochondria with cristae shown clearly in cross section were outlined in red wax pencil, and only these areas were used for calculating cristae surface densities (Smith and Page, 1976). A subset of micrographs of Chionodraco rastrospinosus and Chaenocephalus aceratus mitochondria were also printed, and cristae surface density was quantified from both prints and micrographs projected onto a Summagraphics II digitizing tablet (N=2 per species). Because there was no significant difference between measurements made from projected micrographs and prints in these two species (P=0.92), cristae surface densities were quantified using projected micrographs for the remaining individuals. Micrographs from all three species were overlaid with a square lattice test pattern (d=0.08 μm) for analysis.

Animals were killed and hearts extracted as described above. Assays requiring fresh tissue, for hexokinase (HK), phosphofructokinase (PFK), cytochrome oxidase (CO) and carnitine palmitoyltransferase-I (CPT-I), were performed immediately. For all other assays, pyruvate kinase (PK), lactate dehydrogenase (LDH), citrate synthase (CS) and 3-hydroxyacyl CoA dehydrogenase (HOAD), tissues were quickly frozen in liquid nitrogen, stored at −70 °C and shipped on dry ice to our laboratory at the University of Maine, where they were stored at −70 °C.

For all enzymes other than CO and CPT-I (see below), tissues were homogenized in a 10 % w/v ice-cold buffer (40 mmol l⁻¹ Hepes, 1 mmol l⁻¹ EDTA, 2 mmol l⁻¹ MgCl₂, pH 7.8 at 1 °C). Dithiothreitol (DTT; 2 mmol l⁻¹) was added to the buffer for the PFK, LDH and HK assays. Tissue was homogenized by hand using a ground-glass homogenizer. Homogenates were further reduced by brief (3–5 s) treatment with a Tekmar Tissuemizer and finally homogenized to completion by hand using a ground-glass homogenizer.

All assays were performed in triplicate at 1±0.5 °C using a Perkin Elmer Lambda 6 spectrophotometer. Temperature was maintained using a refrigerated, circulating water bath attached to the spectrophotometer. Background activity was measured in the absence of initiating substrate. Assay conditions are described in detail below. Maximal activities were determined by measuring the rate of oxidation or reduction of pyridine nucleotides at 340 nm for 5 min, except when noted otherwise below.

The methodology employed was slightly modified from that described by Opie and Newsholme (1967) and Read et al. (1977). The final reaction mixture contained 7 mmol l⁻¹ MgCl₂, 200 mmol l⁻¹ KCl, 1 mmol l⁻¹ KCN, 2 mmol l⁻¹ AMP, 0.15 mmol l⁻¹ NADH, 2 mmol l⁻¹ ATP, 4 mmol l⁻¹ fructose 6-phosphate (F6P), 2 units ml⁻¹ aldolase, 10 units ml⁻¹ triosephosphate isomerase, 2 units ml⁻¹ glycerol-3-phosphate dehydrogenase, 75 mmol l⁻¹ triethanolamine, pH 8.4 at 1 °C. Reactions were initiated by the addition of a mixture of ATP and F6P.

The procedure for this assay was that described by Hansen and Sidell (1983). The final reaction mixture contained 2.5 mmol l⁻¹ pyruvate, 0.15 mmol l⁻¹ NADH, 1 mmol l⁻¹ KCN, 50 mmol l⁻¹ imidazole, pH 7.7 at 1 °C. Reactions were initiated by the addition of pyruvate.

The method used for this assay was that described by Hansen and Sidell (1983). The final reaction mixture contained 150 mmol l⁻¹ KCl, 1 mmol l⁻¹ KCN, 10 mmol l⁻¹ MgSO₄, 0.15 mmol l⁻¹ NADH, 5 mmol l⁻¹ ADP, 2.5 mmol l⁻¹ phosphoenolpyruvate (PEP), 10 units ml⁻¹ LDH, 50 mmol l⁻¹ imidazole, pH 7.1 at 1 °C. Reactions were initiated by the addition of PEP.

The protocol for this assay was that originally described by Beenakkers et al. (1967) as modified by Hansen and Sidell (1983). The final reaction mixture contained 1 mmol l⁻¹ EDTA, 1 mmol l⁻¹ KCN, 0.15 mmol l⁻¹ NADH, 0.1 mmol l⁻¹ acetoacetyl CoA, 50 mmol l⁻¹ imidazole, pH 7.7 at 1 °C. Reactions were initiated by the addition of acetoacetyl CoA.

For this assay, we used a modification of the protocol originally described by Srere et al. (1963). The final reaction mixture contained 0.25 mmol l⁻¹ 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB), 0.4 mmol l⁻¹ acetyl CoA, 0.5 mmol l⁻¹ oxaloacetate, 75 mmol l⁻¹ Tris-HCl, pH 8.2 at 1 °C. The reaction was initiated by the addition of oxaloacetate. The progress of the reaction was monitored by following the production of the reduced anion of DTNB at 412 nm.

This assay was modified from that described by Zammit and Newsholme (1976). The final reaction mixture contained 7.5 mmol l⁻¹ MgCl₂, 0.8 mmol l⁻¹ EDTA, 1.5 mmol l⁻¹ KCl, 0.4 mmol l⁻¹ NADP, 2.5 mmol l⁻¹ ATP, 10.0 mmol l⁻¹ creatine phosphate, 1.0 mmol l⁻¹α-D-glucose, 0.9 units ml⁻¹ creatine phosphokinase, 0.7 units ml⁻¹ glucose-6-phosphate dehydrogenase, 75 mmol l⁻¹ Tris-HCl, pH 7.6 at 1 °C. Reactions were initiated by the addition of glucose.

The method of Wharton and Tzagoloff (1967) was used to measure activity. Tissue was homogenized in 50 mmol l⁻¹ K₂HPO₄/KH₂PO₄, 0.05 % Triton X-100, pH 7.5. The assay medium consisted of 10 mmol l⁻¹ K₂HPO₄/KH₂PO₄, 0.65 % (w/v) reduced (Fe²⁺) cytochrome c and 0.93 mmol l⁻¹ K₃Fe(CN)₆. The reaction was initiated by the addition of enzyme. Maximal activities were measured by following the oxidation of reduced cytochrome c at 550 nm.

Maximal activities of CPT-I were measured in intact isolated mitochondria (Rodnick and Sidell, 1994). Tissue was homogenized in 10 % (w/v) of ice-cold 40 mmol l⁻¹ Hepes, 10 mmol l⁻¹ EDTA, 5 mmol l⁻¹ MgCl₂, 150 mmol l⁻¹ KCl, 35 mmol l⁻¹ sucrose, and 0.5 % bovine serum albumin (BSA), pH 7.27 at 1 °C, using a Duall ground-glass homogenizer. A sample of the crude homogenate was reserved for measuring total CPT-I activity. The homogenate was centrifuged at 270 g for 10 min. The supernatant was collected and centrifuged at 270 g. The supernatant was again collected and centrifuged at 15 000 g for 20 min. The mitochondrial pellet was gently resuspended in homogenization buffer (minus BSA) and centrifuged at 15 000 g for 20 min. The resultant pellet was gently resuspended in homogenization buffer lacking BSA to give a final concentration of approximately 5 μg protein μl⁻¹. A sample of the mitochondrial suspension was frozen at −70 °C for later protein determination using the bicinchoninic acid method (Smith et al., 1985).

The final assay medium consisted of 1.0 mmol l⁻¹ EGTA, 220 mmol l⁻¹ sucrose, 40 mmol l⁻¹ KCl, 0.13 % BSA, 0.1 mmol l⁻¹ DTNB, 40 μmol l⁻¹ palmitoleoyl-CoA, 1 mmol l⁻¹ carnitine, 20 mmol l⁻¹ Hepes, pH 8.0 at 1 °C. Activity was simultaneously measured in six cuvettes. Malonyl-CoA, a known inhibitor of CPT-I, was added to three of the six cuvettes to a final concentration of 10 μmol l⁻¹. Reactions were initiated by the addition of carnitine. Maximum activity was measured by following the production of the reduced anion of DTNB at 412 nm. Maximal activities of CPT-I were estimated as the fraction of total activity inhibited in the presence of malonyl-CoA.

Data from transmural sections (V1–V3) were pooled for each individual. Mitochondrial and myofibril volume densities were transformed using an arcsin transformation. Comparisons among the three species of each stereological variable and of the maximal activity of each enzyme were analysed using an analysis of variance (ANOVA) with a post-hoc Fisher’s least-significant-difference test.

Values are presented as means ± S.E.M.

The most striking differences in ultrastructure of the heart among the three species are differences in mitochondrial densities (Fig. 1). Mitochondrial volume densities are highest in Chaenocephalus aceratus, which lacks both Hb and Mb, intermediate in Chionodraco rastrospinosus, which lacks Hb but expresses Mb, and lowest in G. gibberifrons, which expresses both Hb and Mb. Mitochondrial surface densities parallel this trend, with surface densities being highest in Chaenocephalus aceratus, intermediate in Chionodraco rastrospinosus and lowest in G. gibberifrons (Table 1).

The mitochondria are structurally different among the three species. Mitochondrial cristae surface densities, S_v(imm,mit), are higher in G. gibberifrons than in Chaenocephalus aceratus and Chionodraco rastrospinosus. Cristae surface densities tend to be higher in Chionodraco rastrospinosus than in Chaenocephalus aceratus, although there is no statistically significant difference between the two (P=0.36). The surface-to-volume ratio of mitochondria varies among the three species, being higher in G. gibberifrons (7.55±0.26 μm⁻¹) than in Chionodraco rastrospinosus (6.75±0.40 μm⁻¹) and Chaenocephalus aceratus (4.52±0.27 μm⁻¹). Thus, the species that expresses both Hb and Mb has small mitochondria with densely packed cristae. The species that expresses Mb, but not Hb, has slightly larger mitochondria and more loosely packed cristae, and the species that lacks both Hb and Mb has very large mitochondria with a low cristae surface density (Fig. 2).

Mitochondrial cristae surface density per gram of tissue, a generally accepted indicator of aerobic metabolic capacity, is higher in Chaenocephalus aceratus (6.91±0.39 m² g⁻¹) than in G. gibberifrons (4.46±0.31 m² g⁻¹) and Chionodraco rastrospinosus (4.11±0.24 m² g⁻¹) (Table 1). Myofibrillar volume densities are higher in G. gibberifrons than in Chaenocephalus aceratus and Chionodraco rastrospinosus, which may reflect a greater capacity for power output in hearts of G. gibberifrons per volume of tissue than in the other species (Table 1).

Transmural sections of cardiac tissue from each of the three species were subdivided into three regions, and cellular variables were quantified within each region to determine whether they varied among different areas of the heart. These analyses indicate that all the cellular structures measured are distributed homogeneously within the cardiac muscle (data not shown).

Despite significant ultrastructural differences among the three species, differences in mass-specific metabolic indices (per gram of tissue) are minimal. We measured the maximal activity per gram wet mass of several enzymes from different metabolic pathways. Values expressed in this fashion allow us to compare inherent metabolic capacities of cardiac muscle tissue among the different species, despite their differences in heart-to-body-mass ratios. The maximal mass-specific activity of cytochrome oxidase (CO), an indicator of aerobic metabolic capacity, is similar among the three species (Table 2). Because Chaenocephalus aceratus has a significantly higher cristae surface density per gram of tissue than G. gibberifrons and Chionodraco rastrospinosus, this result suggests that electron transport elements may not be as densely packed within the mitochondrial inner membrane of Chaenocephalus aceratus as in the other two species.

The maximal activity of hexokinase (HK) is generally considered a good indicator of the capacity for aerobically oxidizing glucose (Crabtree and Newsholme, 1972a). Mass-specific HK activity is similar among the three species (Table 2). Since the maximal activities of HK, CO and citrate synthase (CS), another aerobically poised enzyme, are similar among the three species, it appears that the absence of oxygen-binding proteins does not compromise the aerobic metabolic capacity of each gram of heart tissue.

CPT-I catalyses a critical step in the translocation of long-chain fatty acids into mitochondria and reflects the capacity for fatty acid oxidation (Crabtree and Newsholme, 1972b). Mass-specific CPT-I activity is higher in the two species that lack oxygen-binding proteins (Chaenocephalus aceratus and Chionodraco rastrospinosus) than in G. gibberifrons. These data also indirectly imply that overall aerobic metabolic capacity may not be compromised in the channicthyids (Table 2).

Anaerobic metabolic capacity indexed by the maximal mass-specific activity of PFK, a key enzyme in the glycolytic pathway, is greatest in hearts of G. gibberifrons among the species examined (Table 2). Thus, despite the loss of expression of Hb and/or Mb, the hearts of channicthyids do not appear to have a greater reliance on anerobic glycolysis to fuel muscular work compared with red-blooded species.

Organismal capacities for cardiac work Chionodraco rastrospinosus have a larger heart-to-body-mass ratio (4.105±0.140 g ventricle kg⁻¹ body mass, N=27) than the other two species examined (Chaenocephalus aceratus 3.255±0.084 g ventricle kg⁻¹ body mass, N=30; G. gibberifrons, 0.715±0.016 g ventricle kg⁻¹ body mass, N=30) (means ± S.E.M.). Expressing maximal enzyme activites per 100 g body mass accounts for these differences and may provide insight about the total metabolic capacity of hearts in vivo and the organismal capacity for cardiac work. When enzymatic activities are expressed in this fashion, Chionodraco rastrospinosus and Chaenocephalus aceratus have the highest capacity for cardiac aerobic metabolism, as indicated by the highest activities of CO, CS and HK (Table 3). Chionodraco rastrospinosus has the greatest capacity for fatty acid oxidation (CPT-I). The capacity for anaerobic glycolysis, as reflected in the maximal activity of PFK, is higher in Chaenocephalus aceratus than in Chionodraco rastrospinosus and G. gibberifrons. Thus, the total metabolic capacity of heart ventricular muscle is greatest in the Channicthyidae, despite their lack of oxygen-binding proteins.

Ultrastructural variables expressed per 100 g body mass indicate that Chionodraco rastrospinosus and Chaenocephalus aceratus have higher mitochondrial volumes and surface areas than G. gibberifrons. Mitochondrial cristae surface areas expressed per 100 g body mass are also highest in Chionodraco rastrospinosus and Chaenocephalus aceratus, as are myofibril volumes (Table 4). These ultrastructural differences correlate with the greater aerobic metabolic capacities per 100 g body mass of the channicthyids.

Our results show a clear correlation between the evolutionary loss of oxygen-binding proteins and substantial differences in the cellular architecture of heart ventricles in Antarctic fishes. The heart of Chaenocephalus aceratus, which lacks both Hb and Mb, has a considerably higher density of mitochondria (37 %) than the heart of Chionodraco rastrospinosus (20 %), which lacks Hb, but whose heart does express Mb. Mitochondrial densities are lowest in hearts from G. gibberifrons (16 %), which expresses both oxygen-binding proteins. By comparing the hearts of Chionodraco rastrospinosus and G. gibberifrons, we can isolate features of the heart correlated with the loss of Hb, and by comparing the hearts of Chaenocephalus aceratus and Chionodraco rastrospinosus, we can examine characteristics correlated specifically with the loss of Mb. Exploiting these comparisons, we conclude that the loss of Hb alone is correlated with only a modest increase in mitochondrial volume density (4 % of cell volume), while the loss of Mb expression correlates with a more substantial increase in the fraction of cell volume occupied by mitochondria (17 % of cell volume). No species of Antarctic fish has been identified that expresses Hb and lacks cardiac Mb. Consequently, we cannot determine whether loss of Mb, in the presence of Hb, would result in a similar expansion of mitochondrial density. Thus, we cannot rule out the possibility that ultrastructural alterations in the heart of Chaenocephalus aceratus may be due to the combined effects of the loss of both Hb and Mb that may be greater in magnitude than the additive effects of losing either protein separately.

During periods of intense activity, blood levels may decline more precipitously in hearts of icefish than in red-blooded fishes because of the diminished oxygen-carrying capacity of their blood (Ruud, 1954). Oxygen delivery to mitochondria may be further constrained in species that lack Mb, which serves both to facilitate oxygen diffusion and as an intracellular reservoir of oxygen.

The architecture of cardiac myocytes in channicthyids may compensate for the loss of respiratory proteins and contribute to maintaining oxygen diffusion to mitochondria. All three species studied have a type I heart, lacking a coronary circulation (Davie and Farrell, 1991). Oxygen utilized by respiring mitochondria must diffuse from the mixed-venous blood present in the lumen of the heart. Because mitochondria are randomly distributed within the ventricle of each species, as mitochondrial density increases, the diffusion distance between the cell surface and the mitochondrial membrane decreases, effectively reducing X in the equation above (K. M. O’Brien and B. D. Sidell, unpublished results). In addition, there may be differences among the three species in the degree of trabeculation of the spongy myocardium. The mean diffusion distance between the ventricular lumen and the mitochondrial membrane may be reduced in hearts from species lacking oxygen-binding proteins if they are more highly trabeculated than hearts from fishes that express one or both of these proteins. We are currently testing this hypothesis using a stereologically based model developed for quantifying mean oxygen diffusion distance within hearts from each of the three species.

High mitochondrial densities within tissues also provide a network of lipid-rich intracellular membranes that may act as conduits for oxygen movement. Oxygen is more than four times more soluble in non-polar solvents than in water (Battino et al., 1968). The resultant higher solubility constant of oxygen within lipid-rich membranes compared with aqueous cytoplasm may enhance the rate of transcellular oxygen diffusion in mitochondria-rich cells. The importance of intracellular lipids in enhancing oxygen movement has been recognized in other fishes (Egginton and Sidell, 1989; Londraville and Sidell, 1990). Oxidative skeletal muscles from striped bass accumulate high densities of intracellular lipid droplets in response to cold-temperature acclimation (Egginton and Sidell, 1989). Subsequent measurements showed that these increases in the density of intracellular lipid droplets result in a significant increase in the solubility constant of oxygen, leading to an enhanced intracellular rate of oxygen diffusion (Desaulniers et al., 1996). Although increases in membrane densities were not accounted for in this study, several others have highlighted the potential importance of intracellular membranes in oxygen transport.

Longmuir (1980) reported that oxygen is transported more rapidly between blood and mitochondria along channels of high solubility than through the aqueous cytoplasm. He hypothesized that the endoplasmic reticulum accounted for these ‘channels’ of oxygen movement. Mitochondrial membranes may serve as similar conduits for oxygen diffusion. The properties of the lipid bilayer of a membrane appear to determine its effectiveness in transporting oxygen. Experiments on isolated mitochondrial and plasma membranes from bullfrog heart tissue show that mitochondrial membranes are less viscous than plasma membranes, resulting in a higher diffusion coefficient for oxygen and an enhanced rate of oxygen diffusion. The lower viscosity of mitochondrial membranes compared with plasma membranes is due to an increased proportion of unsaturated acyl chains within the constituent phospholipids of the mitochondrial bilayers (Koyama et al., 1990).

The organization of intracellular membranes may be as important as lipid composition in determining their effectiveness as pathways for oxygen transport. Mitochondria forming a continuous reticulum within a cell may provide the best conduit for oxygen diffusion because the pathway for oxygen diffusion is continuous (Dutta and Popel, 1995). This would result in a higher diffusive flux compared with that in a cell containing mitochondria separated by aqueous cytoplasm, requiring oxygen to diffuse across a heterogeneous path of both lipid and cytoplasm. The efficiency of oxygen transport within a membranous network may explain why hearts from Chaenocephalus aceratus possess such a high density of large mitochondria. These enlarged mitochondria enable a juxtaposition of the outer mitochondrial membranes. Because the outer mitochondrial membranes are less protein-dense than the inner membrane cristae, they will provide the best membranous pathway for oxygen diffusion and may compensate for the absence of Mb.

Under some conditions, intracellular lipids may be more effective than myoglobin at transporting oxygen. More oxygen is found dissolved in lipid droplets within the skeletal muscle of cold-acclimated striped bass than bound to myoglobin (Desaulniers et al., 1996). There are also notable differences in the behavior of oxygen dissolved in lipid compared with oxygen bound as a ligand to Mb. Oxygen present within intracellular lipid is able to move freely from regions of high to regions of low ⁠. In contrast, oxygen bound to Mb dissociates from the protein only at very low levels. Therefore, lipid may be more critical than Mb for ensuring adequate oxygen delivery within tissue at normal activity levels, and Mb may play a backup role, releasing oxygen only during strenuous activity (Sidell, 1998).

Mitochondrial volume density is normally indicative of the oxidative capacity of a tissue: high mitochondrial density typically reflects high metabolic demand. Hummingbirds have the highest mass-specific metabolic rates among vertebrates and also possess nearly the highest mitochondrial densities found in muscle (37 %) (Suarez et al., 1991). Thus, it may be somewhat surprising to observe comparable mitochondrial densities in the heart of Chaenocephalus aceratus, which lacks Hb and Mb and is a sluggish, demersal species. Closer examination of significant differences in the architecture of mitochondria among the three species, however, may explain this apparent anomaly.

Mitochondrial cristae density is also usually positively correlated with respiration rate and oxidative capacity (Schwerzmann et al., 1989). Cristae surface densities vary among the three species and are inversely proportional to mitochondrial volume densities. Cristae are more densely packed within the mitochondria of hearts of G. gibberifrons than in those of Chionodraco rastrospinosus and Chaenocephalus aceratus. The lower cristae surface densities within mitochondria suggest that icefish might have a lower oxidative capacity than the red-blooded species. However, when the densities of inner-mitochondrial membranes are calculated per gram of tissue, the heart from Chaenocephalus aceratus has a significantly greater cristae surface area than those from the other two species, whose cristae areas are equivalent. These results suggest that the oxidative capacity per gram of ventricle from the myoglobinless icefish Chaenocephalus aceratus might be greater than that of the heart from the two species that express oxygen-binding proteins.

To gain a better insight into the aerobic capacity of hearts from all three species, we also measured the maximal activites of several aerobically poised enzymes. The activity of CO is usually proportional to respiration rate and to the surface density of inner-mitochondrial membrane per gram of tissue. Our results show that the maximal activity of cytochrome c oxidase (CO) per gram of tissue is equivalent among hearts from all three species. The apparent mismatch between cristae surface density per gram of tissue and CO activity within the heart of Chaenocephalus aceratus may be reconciled if the electron transport elements are less densely packed within the inner mitochondrial membranes than in the other two species. Alternatively, there may be differences in the catalytic rate constant (k_cat) of CO among the three species. Because all three species are closely related in phylogeny, however, it seems unlikely that they would express markedly different variants of CO. This does not, however, rule out differences in the lipid composition of the mitochondrial membranes among these fishes that may also affect the catalytic capacity of CO.

In addition to the activity of CO, the maximal activities of other enzymes from aerobic pathways (HK, CS) are also equivalent on a mass-specific basis among hearts from the three species, indicating that aerobic metabolic capacity is not diminished in the absence of oxygen-binding proteins. Similar results were reported by Driedzic and Stewart (1982), who found no differences in the maximal activities of CO, CS and HK between hearts from the Atlantic ocean pout Macrozoarces americanus, which lacks Mb, and the sea raven Hemitripterus americanus, which expresses the protein. No information is available for these species to evaluate whether ultrastructural differences in cardiac muscle between the species might maintain oxygen delivery to the mitochondria and aerobic metabolic rates.

The activities of enzymes from pathways of fatty acid oxidation (CPT-I, HOAD) are greatest in hearts from channicthyids. In addition, hearts from G. gibberifrons have a higher activity of PFK compared with icefishes. These data provide further evidence not only that aerobic metabolic capacity is not compromised in species lacking oxygen-binding proteins but also that hearts from these species do not appear to rely more on anerobic pathways to fuel heart work than those of their red-blooded relatives.

In summary, the metabolic characteristics of the three species examined were remarkably similar despite differences in the expression of oxygen-binding proteins. We did, however, observe striking differences in cellular architecture correlated with the expression of oxygen-binding proteins. The high densities of mitochondria within hearts of species that lack Hb and/or Mb may contribute to maintaining oxygen flux to mitochondria by two mechanisms. First, high mitochondrial densities shorten the diffusion distance between the lumen of the heart and the mitochondrial membrane. Second, the membranous network created by large mitochondrial densities provides a favorable pathway for oxygen movement because of the higher solubility of oxygen in lipid than in cytoplasm. Structural alterations in the cardiac myocytes of Antarctic fishes that lack oxygen-binding proteins may therefore overcome potential reductions in oxygen diffusion rates so that aerobic metabolic capacities are equivalent to those of fishes that express hemoglobin and/or myoglobin.

We greatly appreciate the excellent support from the personnel at the US Antarctic research station, Palmer Station, and the masters and crew of R/V Polar Duke. Funding for this study was provided by US National Science Foundation Grants OPP 92-20775 and OPP 94-21657 to B.D.S.