AMP-activated protein kinase activity during metabolic rate depression in the hypoxic goldfish, Carassius auratus

Environmental hypoxia is a common, naturally occurring phenomenon in many aquatic ecosystems, the prevalence of which is increasing due to anthropogenic nutrient loading and eutrophication of both freshwater and marine environments. The major metabolic challenge facing animals living in hypoxic environments is the inhibition of ATP production via oxidative phosphorylation, thus hypoxia-intolerant animals quickly succumb to hypoxia due to an inability to maintain cellular energy balance and a loss of cellular[ATP] (Boutilier, 2001). By contrast, hypoxia-tolerant animals are able to maintain cellular energy balance when faced with an O₂ limitation by reducing metabolic demands to match the limited capacity for O₂-independent ATP production. At the cellular level, metabolic rate depression is achieved by reducing rates of protein synthesis and membrane ion movement among other processes (Hochachka et al.,1996). In hepatocytes isolated from the anoxia-tolerant turtle(Chrysemys picta belli) ATP-utilization during anoxia exposure falls to ∼10% of normoxic rates (Buck et al., 1993) with reductions in protein synthesis making up the largest percentage of this decline (Land et al., 1993). In addition, similar hypoxia-induced reductions in protein synthesis rates have been observed in the anoxia-tolerant crucian carp Carassius carassius (Smith et al., 1996). Hypoxia survival necessitates the timely and well-synchronized restructuring of cellular processes involved in both energy provision and utilization; however, little research has focused on how these cellular processes are co-ordinated during hypoxia in order to maintain cellular energy balance.

AMP-activated protein kinase (AMPK) represents an ideal candidate protein to co-ordinate the metabolic responses to hypoxia. AMPK is a heterotrimeric protein kinase composed of a catalytic subunit (α) and two regulatory subunits [β and γ (Carling,2004)]. Phosphorylation of AMPK at Thr-172 on the α-subunit is essential for its activation (Carling,2004) and this is brought about via the activity of upstream kinases. Two of these upstream kinases have been identified, LKB1(Sakamoto et al., 2005) and CaMKK (Witters et al., 2006),in mammals. AMPK appears to be continuously phosphorylated; however, the phosphate group is rapidly removed under normal conditions returning AMPK to an inactive form (Hardie,2007). Binding of AMP to AMPK induces a conformational change and prevents dephosphorylation (Sanders et al., 2007). Upon activation, AMPK inhibits anabolic processes in the cell and activates catabolic processes(Hardie et al., 2006), thereby helping to maintain cellular [ATP]. In mammalian models, AMPK has been shown to inhibit protein synthesis through phosphorylation of eukaryotic elongation factor 2 [eEF2 (Horman et al.,2002)], decrease glycogen synthesis rates through inactivation of glycogen synthase (Nielsen et al.,2002) and decrease fatty acid synthesis rates through phosphorylation of acetyl-CoA carboxylase-1 [ACC-1(Hardie and Pan, 2002)]. Activation of AMPK has also been shown to result in increased skeletal muscle hexokinase activity, glucose transporter expression [GLUT-4(Holmes et al., 1999)] and translocation to the membrane(Kurth-Kraczek et al., 1999),and increased phosphofructokinase-2 (PFK-2) activity in rat cardiomyocytes(Marsin et al., 2000), all of which could enhance O₂-independent ATP production. Combined, these actions have led to AMPK being termed the cellular `energy gauge' because of it critical role in maintaining cellular energy balance. However, the cellular consequences of AMPK activation have been studied mainly in exercise- and ischemia-stressed mammalian models, and no study has assessed the role of AMPK in co-ordinating the cellular responses to environmental hypoxia in a hypoxia-tolerant fish.

At the extreme of hypoxia-tolerance among teleost fishes are the Carassius sp., which are capable of surviving months of anoxia at cold temperature. An important means by which members of this genus accomplish this feat is through a strong hypoxia-dependant depression of metabolic rate and activation of substrate-level phosphorylation. This has been described in the common goldfish, Carassius auratus, which depresses metabolic rate by ∼70% during anoxic bouts [assessed via direct calorimetry(Van Waversveld et al.,1989)]. Metabolic depression during hypoxia/anoxia is key for goldfish survival as it allows for the conservation of endogenous glycogen reserves thereby extending the amount of time that can be spent under O₂-limiting conditions.

Given that AMPK is sensitive to changes in cellular energy status and that its activation leads to a general reduction in anabolic pathways and a stimulation of catabolic pathways, we hypothesize that it may play a role in co-ordinating the processes involved in the metabolic rate depression observed in the goldfish during exposure to severe hypoxia. In the present study, we determined cellular energy status, activation pattern of AMPK and its interactions with a well-characterized target, eEF2, and protein synthesis in liver and skeletal muscle of normoxia- and hypoxia-exposed goldfish. This was carried out in an attempt to determine whether or not AMPK may play a role in co-ordinating metabolic depression during hypoxia exposure in hypoxia-tolerant organisms.

Adult goldfish (Carassius auratus L.) weighing 36.0±1.4 g(means ± s.e.m.) were purchased from a local supplier (Delta Aquatics,Richmond, BC, Canada) and held in either flow-through or static renewal dechlorinated City of Vancouver tap water at 16°C. Fish were fed daily with commercial goldfish flakes. All animal procedures adhere to the Canadian Council on Animal Care guidelines as administered by the University of British Columbia Animal Care Committee.

Goldfish were sampled directly from a holding tank and sacrificed with an overdose of benzocaine (1 gl^–1). Samples of brain, eye,heart, gill, intestine, liver, kidney and muscle were rapidly excised,flash-frozen in liquid nitrogen, and stored at –80°C. Total RNA was extracted from these tissue samples following the methods of Chomczynski(Chomczynski, 1993) using Tri Reagent (Sigma-Chemical Co., St Louis, MO, USA). Following isolation, total RNA was quantified spectrophotometrically and the integrity of the two ribosomal bands was assessed by electrophoresis. RNA was stored at–80°C. Reverse transcription reactions and PCR amplification of AMPK sequences were carried out following the methods outlined in Richards et al.(Richards et al., 2003). Briefly, cDNA was synthesized from 4 μg total RNA using RevertAid H Minus M-MuLV Reverse Transcriptase (Fermentas, Burlington, ON, Canada) following the manufacturer's instructions. Partial AMPK subunit sequences were obtained using primers designed from the conserved regions of known AMPK subunit isoforms (α1, α2, β1, β2, γ1, γ2, andγ3) using all available vertebrate sequence information in GenBank,although only primers designed for α1 and β1 yielded amplicons. Primers for AMPKα1a were (forward) 5′-GGG CCA GCG TAA AAC CTT CCT-3′ and (reverse) 5′-GGA GGG GAA CTG TTT GAT TAT AT-3′,and PCR for this gene product consisted of 35 cycles; 1 min at 94°C, 1 min at 51°C and 2 min at 72°C. Primers for AMPKα1b were (forward)5′-GGA GGG GAG CTA TTT GAT TAT AT-3′ and (reverse) 5′-GGG TTC TTC TTC GTA CAC G-3′, and PCR for this gene product consisted of 35 cycles; 1 min at 94°C, 1 min at 53°C and 2 min at 72°C. Primers for AMPKβ1 were (forward) 5′-GCC GGA AGG AGA GCA TCA GTA CAA GT-3′ and (reverse) 5′-GCG CTA AGA ACC ATC ACG CCA T-3′, and PCR for this gene product consisted of 35 cycles; 1 min at 94°C, 1 min at 60°C and 2 min at 72°C. Primers were designed using GeneTool Lite software(www.biotool.com). PCR products were gel purified and ligated into a plasmid vector (pGEM-T EasyVector System II; Promega, Madison, WI, USA). Ligated plasmids were transformed into heat-shock competent Escherichia coli (strain JM109;Promega) and plated onto LB-agar plates. Colonies were grown overnight at 37°C and several colonies containing the ligated insert were selected and grown in liquid culture. Following overnight culture, plasmid DNA was harvested from cultured cells using a GenElute Plasmid Miniprep kit (Sigma Chemical Co.) and sequenced on an Applied Biosystems PRISM 377 sequencer(Foster City, CA, USA).

Tissue distribution of goldfish AMPK isoforms was estimated using quantitative real-time PCR (qPCR) and isoform-specific primers designed using Primer Express software (Applied Biosystems, Foster City, CA, USA). Primers for AMPKα1a (GenBank accession number EU583380) were (forward)5′-GCC AAG ATC GCT GAC TTT GG-3′ and (reverse) 5′-CGC AGC TCG TTC TCA GGA A-3′. Primers for AMPKα1b (EU583381) were(forward) 5′-TAA GGA CGA GTT GCG GTT CTC-3′ and (reverse)5′-GCC CTG CGT ATA ACC TTC CA-3′. Primers for AMPKβ1(EU580137) were (forward) 5′-GCT GCA GGT GCT CCT CAA C-3′ and(reverse) 5′-GTT GAG CAT CAC ATG GGT TGG T-3′. Total RNA was extracted from brain, eye, heart, gill, intestine, liver, kidney and muscle from fish sampled directly from the holding tank and cDNA was prepared using the same methods as outlined above. Expression was quantified by qPCR using an ABI PRISM 7000 sequence detector (Applied Biosystems). qPCR reactions consisted of 2 μl cDNA (reverse transcribed from 4 μg or total RNA), 4 pmol of each primer and Universal SYBR green master mix (Applied Biosystems)in a total volume of 22 μl. qPCR conditions included initial incubations of 2 min at 50°C and 10 min 95°C, followed by 40 cycles consisting of 15 s at 95°C and 1 min at 60°C. Melt-curve analysis was performed following each reaction to ensure that only a single product was amplified. Additionally, random products were sequenced following the methods outlined above to ensure the amplified product was indeed the product of interest.

Three weeks before experimentation, a group of ∼80 fish were transferred into a 375 l aquarium equipped with a canister filter and a cooling coil. Water temperature was then lowered in the tank at a rate of 2°C per day using a re-circulating water-chiller until it reached 10°C, at which point temperature was maintained for at least two weeks prior to experimentation. Fish were fed commercial goldfish flakes daily throughout the acclimation period.

Thirty-six hours before experimentation, goldfish were transferred into individual exposure chambers and returned to the aquarium. The exposure chambers consisted of highly perforated plastic beakers that allow for good water exchange between the exposure chamber and the bulk water and were large enough so the fish could move freely. These chambers were designed so that they slid smoothly into basins that were slightly larger than the exposure chamber and we could remove the fish from the aquarium without air exposure or causing agitation. An overdose of benzocaine (1 gl^–1) could then be added to the basin and the fish sampled. To obtain normoxic tissue samples, eight chambers, each containing a fish, were removed and benzocaine added. At complete anaesthesia, which occurred ∼3 min following the addition of benzocaine, individual fish were removed, patted dry, and weighted to the nearest 0.1 g. Blood was sampled following caudal severance using haematocrit (Ht) tubes and samples of skeletal muscle, liver, heart, brain and gill were rapidly excised, flash-frozen in liquid nitrogen, and stored at–80°C.

It should be noted that many anaesthetics are known to affect protein phosphorylation [e.g. tetracaine (Nivarthi et al., 1997)] but nothing is known of the impacts of benzocaine on protein phosphorylation; however, since all fish in the present study were sampled in an identical manner, any changes in protein phosphorylation observed are due to hypoxia treatment and not the anaesthetic chosen.

Following the sampling of normoxic fish, the water [O₂] in the experimental tank was lowered over a 1 h period by bubbling nitrogen-gas into the water, until it reached ∼0.3 mg l^–1. Water[O₂] was monitored throughout the course of hypoxia exposure using an Oakton DO 6 dissolved O₂ meter (Cole Parmer, Montreal, QC,Canada). Eight fish were sacrificed at each of the six time points (0.5, 1, 2,4, 8 and 12 h hypoxia exposure) in an identical manner to normoxic fish. Water temperature was maintained at 10°C throughout the experiment.

To obtain sufficient tissue for complete biochemical analysis, the acclimation and experimental trials were performed twice. Fish from the first experiment were used for the determination of muscle and liver intracellular pH (pH_i), muscle metabolite concentrations, and muscle and liver AMPK activity, protein content and mRNA expression levels. Fish from the second experiment were used for the determination of haematology, plasma[lactate], liver pH_i, metabolites, eEF2 and phospho-Thr-56 eEF2 protein expression, and analysis of liver protein synthesis rates. Liver pH_i and [lactate] were determined in both experiments and no significant differences were found between the two experiments [data not shown; two-way analysis of variance (ANOVA), P>0.05], therefore we consider both experiments to be comparable.

Blood [haemoglobin] (Hb) was determined spectrophotometrically using Drabkin's reagent (Blaxhall and Daisley,1973). Haematocrit was determined by centrifugation of whole blood at 5000 g for 3 min in sealed capillary tubes. Mean cellular haemoglobin content (MCHC) was calculated as [Hb]/Ht.

Frozen muscle (∼200 mg) was ground to a fine powder under liquid nitrogen and pH_i was determined in an aliquot following the methods of Pörtner et al. (Pörtner et al., 1991) using a thermostatted Radiometer BMS3 Mk2 capillary microelectrode with PHM84 pH meter (Radiometer, Copenhagen, Denmark). The remaining ground muscle tissue was lyophilized for 72 h and stored above desiccant at –80°C. For pH_i determination in liver,∼50 mg of liver was sonicated using a micro-sonicator (Kontes, Vineland,NJ, USA) at medium frequency for ∼3 s in 0.2 ml ice-cold metabolic inhibitor (Pörtner et al.,1991). Liver pH_i was measured using an ultra-fine Accumet pH electrode (Cole Parmer).

For metabolite determination, ∼20 mg lyophilized skeletal muscle or∼100 mg of frozen liver was homogenized at maximum speed in ice-cold 8%perchloric acid for 30 s using a Polytron homogenizer (Kinematica Inc.,Bohemia, NY, USA). Homogenates were then centrifuged at 20,000 g for 5 min at 4°C and the supernatant adjusted to∼pH7.6 with 3 mol potassium carbonate. Neutralized extracts were centrifuged at 20,000 g for 5 min at 4°C and the supernatant was immediately frozen in liquid nitrogen and stored at–80°C until use. These extracts were then used for the enzymatic determination of tissue [lactate], [ATP] and [creatine phosphate] (CrP)(Bergmeyer, 1983). Total[creatine] (Cr) was determined by heating an aliquot of extract in sealed Eppendorf tubes for 20 min at 60°C and assaying for Cr enzymatically(Bergmeyer, 1983). Free [Cr]was calculated for each sample by subtracting [CrP] from total [Cr]. Plasma[lactate] was measured enzymatically on deproteinized plasma [20 μl 8%perchloric acid added to 20 μl of plasma].

Sample preparation, SDS-PAGE and western blotting were carried out according to the methods outlined by Todgham et al.(Todgham et al., 2005). Briefly, liver and muscle samples (∼20 mg) were homogenized in a buffer containing: 100 mmol l^–1 Tris-HCl; 1% sodium dodecyl sulphate(SDS); 5 mmol l^–1 ethylenediaminetetraacetic acid; 1μgml^–1 aprotinin; 1 μgml^–1 pepstatin A; 1 μgml^–1 leupeptin; 20 μgml¹phenylmethanesulphonylfluoride; pH7.5. Homogenates were centrifuged at 5000 g for 10 min at 4°C, the supernatant was assayed for total protein using the methods of Bradford(Bradford, 1976) and a portion of the supernatant was denatured by boiling it for 3 min in SDS-sample buffer(Laemmli, 1970). Denaturing SDS-polyacrylamide gels were loaded with denatured liver and muscle homogenates at a protein concentration of 20 μg protein per lane and electrophoresed for 15 min at 75 V followed by 75 min at 150 V. An identical control sample was included on each gel to control for gel-to-gel variation. Following electrophoresis, proteins were transferred to nitrocellulose membranes (Bio-Rad Laboratories, Hercules, CA, USA) using a Trans-Blot semi-dry transfer cell (Bio-Rad). Blots for total AMPKα were blocked using Tween-20 Tris-buffered saline [TTBS: 17.4 mmol l^–1Tris-HCl; 2.64 mmol l^–1 Tris Base; 0.5M sodium chloride(NaCl); and 0.05% Tween-20 (v/v)] with 2% (w/v) non-fat powdered milk. Blots for eEF2 and phospho-Thr-56 eEF2 were blocked using TTBS with 3% (w/v) bovine serum albumin. All membranes were incubated overnight at 4°C in a 1:1000 dilution of primary antibody [either rabbit IgG anti-AMPKα, rabbit IgG anti-eEF2 or rabbit IgG anti-phospho-Thr-56 eEF2 (Cell-Signalling Technology,Danvers, MA, USA)]. Following washing in TTBS, membranes were incubated in 1:5000 IgG goat anti-rabbit secondary antibody [alkaline phosphatase conjugated (Sigma Chemical Co.)] in TTBS for 1 h. Membranes were developed in alkaline phosphatase buffer containing 5-bromo-4-chloror-3-indolyl phosphate(BCIP) and nitroblue tetrazolium (NBT; Sigma Chemical Co.). Band intensity was quantified using a FluorChem 8800 imager (Alpha Innotech, San Leandro, CA,USA) assisted by AlphaEase FC software (v. 3.1.2; Alpha Innotech), and protein amount was expressed relative to total homogenate protein loaded into each well and normalized to the normoxic control samples.

AMPK activity was determined following the methods described by Davies et al. (Davies et al., 1989). Briefly, ∼150 mg frozen tissue (muscle or liver) was homogenized for 30 s at medium speed in approximately 3 volumes of ice-cold homogenization buffer[50 mmol l^–1 Tris-base; 250 mmol l^–1Mannitol; 1 mmol l^–1 EGTA; 1 mmol l^–1 EDTA;50 mmol l^–1 sodium fluoride (NaF); 5 mmol l^–1 sodium pyrophosphate; 1 mmol l^–1phenylmethanesulphonyl fluoride (PMSF); 4 g ml^–1 trypsin inhibitor; 1 mmol l^–1 benzamidine; and 1 mmol l^–1 diothioreitol (DTT)]. Samples were then centrifuged at 4°C for 20 min at 14,000 g and 360 l supernatant was transferred to a new micro-centrifuge tube, and 40 l of 25% (w/v) polyethylene glycol-6000 (PEG-6000) was added bringing the concentration in the tube to 2.5% PEG-6000. Sample tubes were then vortexed for 10 min at 4°C and subsequently centrifuged at 10,000 g for 10 min at 4°C. Following centrifugation, 320 l supernatant was transferred to a new micro-centrifuge tube and ∼60 l of 25% PEG-6000 was added bringing the concentration in the tube to 6% PEG-6000. Tubes were again vortexed for 10 min at 4°C and centrifuged at 10,000 g for 10 min at 4°C. The supernatant was then removed and discarded and the pellet was washed with 300 l of 6% PEG-6000 (prepared in homogenization buffer) before being centrifuged for a final time at 10,000 g for 10 min at 4°C. Following centrifugation, the supernatant was removed and discarded and the pellet was resupended in 75 l ice-cold resuspension buffer (50 mmol l^–1 Tris-base; 250 mmol l^–1 Mannitol; 1 mmol l^–1 EGTA; 1 mmol l^–1 EDTA; 50 mmol l^–1 NaF; 5 mmol l^–1 sodium pyrophosphate;10% w/v glycerol; 0.02% sodium azide; 1 mmol l^–1 PMSF; 4 mg ml^–1 trypsin inhibitor; 1 mmol l^–1benzamidine; 1 mmol l^–1 DTT). An aliquot of the purified resuspended protein solution was taken to determine the total protein by the Bradford protein assay [Sigma Chemical Co.(Bradford, 1976)]. Aliquots of 50 l of 1 mg ml^–1 resuspended protein were prepared for each sample in 0.12% Triton X-100 (Sigma Chemical Co.) made up in resuspension buffer and immediately frozen at –80°C for no longer than two weeks before the activity assays were run. At the time of assay, samples were thawed on ice, and 2.5 l of suspension was assayed for total AMPK activity in a final volume of 25 l, containing: 40 mmol l^–1 Hepes; 80 mmol l^–1 NaCl; 8% w/v glycerol; 0.8 mmol l^–1EDTA; 0.2 mmol l^–1 SAMS peptide (GenScript, Piscataway, NJ,USA); 0.2 mmol l^–1 AMP; 0.8 mmol l^–1 DTT;200 mol ATP; 5 mmol l^–1 magnesium chloride; and[³²P]-ATP (∼3500 cpm pmol^–1). Negative controls, where sample was replaced with distilled H₂O, were also run for each sample. After incubation for 5 min at 20°C, 15 l aliquots were spotted onto 2 cm round phosphocellulase paper (Whatman p81, GE Healthcare, Baie d'Urfé, Quebec, Canada) and the phosphorylation reaction immediately stopped by submergence of the spotted papers into 200 ml of 150 mmol l^–1 phosphoric acid. Spotted papers were washed 10 times for 5 min each in the same volume of fresh 150 mmol l^–1 phosphoric acid. Ten washes were necessary to reduce non-specific binding to near background levels. Spotted papers were then washed once in 300 ml of acetone for 5 min and air-dried. The amount of bound ³²P on the papers was assessed using scintillation counting. AMPK activity was initially expressed as nmol of incorporated ³²P min^–1 mg^–1 of total protein; however, we did not run T=0 assays to correct for non-specific radioactivity coming down with the protein (although this binding will be consistent across all samples assayed), therefore, we present AMPK activity relative to the normoxia control sample, which was set to T=1.

The expression of AMPKα1a, AMPKα1b and AMPKβ1 mRNA in liver was estimated using qPCR and cDNA synthesized from extracted mRNA using the above protocols. The qPCR primers were identical to those described above and, in this case, the expression of each gene was normalized against the expression of β-actin. qPCR primers were designed for β-actin using available goldfish sequence (Accession No. AB039726) and were (forward)5′-TGA CCG AGC GTG GCT ACA G-3′ and (reverse) 5′-TCT CCT TGA TGT CAC GGA CAA T-3′. There was no effect of hypoxia exposure on the expression of actin when expressed as a function of total RNA, thus actin appears to be a good control gene for hypoxia studies. To determine the extent of genomic DNA contamination, we developed non-reverse transcribed controls for a random selection of samples. To develop non-reverse transcribed controls, we diluted our RNA samples (containing genomic DNA) to the same extent as our samples used for cDNA synthesis; however, did not reverse transcribe the samples. These samples, along with their paired cDNA sample,were subjected to qPCR and any amplification in the non-reverse transcribed control was due to genomic DNA contamination. Genomic DNA contamination was present in all samples but never constituted more than 1:1024 starting copies for AMPKα1a, 1:32 starting copies for AMPKα1b or 1:524 288 starting copies for AMPKβ1. Genomic DNA, therefore, represents a minor contribution to the total qPCR signal. One randomly selected control sample was used to develop a standard curve relating threshold cycle to cDNA amount for each primer set to assess efficiency of the reaction. All results were expressed relative to these standard curves, and mRNA amounts (in arbitrary units) were normalized to the expression of actin. Expression levels in hypoxia-exposed animals were expressed relative to the mean expression levels in the normoxia control samples. All samples were run in duplicate and the coefficient of variation between duplicate samples was always <10%.

Protein synthesis rates were determined following the methods outlined by Rider et al. (Rider et al.,2006). Briefly, frozen liver was homogenized at 1:5 (w/v) in ice-cold extraction buffer containing: 50 mmol l^–1 Hepes (pH 7.4); 250 mmol l^–1 sucrose; 20 mmol l^–1 NaF;5 mmol l^–1 sodium pyrophosphate; 1 mmol l^–1EDTA; and 1 mmol l^–1 EGTA, and then clarified by centrifugation at 14,000 g for 15 min at 4°C. The resulting supernatant was removed and stored at –80°C until analysis, which was performed within two weeks of extraction. On the day of analysis, Sephadex G-25 columns (GE Healthcare, Piscataway, NJ, USA) were equilibrated with buffer containing: 50 mmol l^–1 Hepes (pH 7.4); 200 mmol l^–1 potassium acetate; 5 mmol l^–1 magnesium acetate; 1 mmol l^–1 DTT; 5 g ml^–1 leupeptin; 1 mmol l^–1 benzamidine; and 1 mmol l^–1 PMSF as instructed by the column manufacturer. Clarified tissue extracts were thawed on ice and 0.5 ml was gravity filtered through columns to remove endogenous amino acids. Filtrate, containing cellular proteins, was collected and analysed for total protein using the Bradford assay as described above. To determine protein synthesis rates, a 50 l aliquot of the filtrate was added to assay buffer containing, 50 mmol l^–1 Mops (pH 7.1), 140 mmol l^–1 potassium acetate, 20 mmol l^–1 magnesium acetate, 2 mmol l^–1 DTT, 20 mmol l^–1 CrP, 20 mmol l^–1 creatine kinase (CrK), 1 mmol l^–1 ATP,0.5 mmol l^–1 GTP, 0.1 mmol l^–1 spermidine,10 U RNaseOUT (Invitrogen, Burlington, ON, Canada), 50 ug ml^–1 total RNA prepared from goldfish liver using the Tri-Reagent method as outlined above (Sigma Chemical Co.), and 20 mmol l^–1 of each amino acid (except leucine) to a final volume of 100 μl. The reaction was started with the addition of 0.9 ul of 20 μmol activated leucine stock containing L-[4,5-³H]-leucine (∼300 cpm pmol^–1) and incubated at 25°C for 90 min. Negative controls, where clarified extract was replaced with distilled H₂O,were assayed for each sample. Following incubation the reaction was immediately stopped with the addition of 1 ml 10% (w/v) trichloroacetic acid and placed on ice for 10 min. Samples were then centrifuged at 10,000 g for 5 min to collect precipitated proteins and the pellet was resuspended in 0.2 ml of 0.1 mol l^–1 sodium hydroxide and re-precipitated in 1 ml of 5% (w/v) trichloroacetic acid. After 10 min on ice,proteins were collected by centrifugation and subjected to an additional wash. Following this wash, proteins were solubilized in 1 ml formic acid and 0.9 ml of the solubilized protein solution was taken for counting in 10 ml of Toluene-based scintillant on a LS 1801 liquid scintillation counter (Beckman Coulter, Mississauga, ON, Canada). Protein synthesis rates are expressed as pmol of leucine incorporated per mg of total protein per hour.

Free cytosolic [ADP] and [AMP] were calculated from measured values of[ATP], [CrP], [Cr] and pH_i assuming equilibrium of the creatine kinase and adenylate kinase reactions. Before calculating [ADP_free]and [AMP_free], metabolite concentrations were converted to molar concentrations using a tissue–water content of 0.8 ml g^–1 wet mass (Wang et al., 1994). The equilibrium constants for creatine kinase(K′_CK) and adenylate kinase(K′_AK) were corrected for experimental temperature,ioinic strength, measured pH and free Mg²⁺ [assumed to be 1 mmol l^–1 (Van Waarde et al.,1990)] according to published protocols(Golding et al., 1995; Teague et al., 1996).

All data are presented as means ± s.e.m. All muscle metabolite concentrations determined on lyophilized tissue were converted back into wet mass using a 4:1 wet:dry ratio. Statistical analysis involved one-way ANOVA followed by Holm–Sidak post hoc test to identify where statistical difference occurred. All data were tested for normality(Kolmogorov–Smirnov test) and homogeneity of variance (Levene median test). In cases where data sets did not meet these assumptions, data were log transformed and statistical analyses repeated. For those data sets that still did not meet assumptions following transformation, statistical analysis involved Kruskal–Wallis one-way ANOVA on ranks followed by Dunn's post hoc test. Differences were considered statistically significant at P<0.05.

We identified two genes coding for AMPKα1 and one gene coding for AMPKβ1 in goldfish. Alignment of the cDNA sequences with the DNA sequences deposited in GenBank revealed homologies of 69–95% for the AMPKα1-catalytic subunit and 78–84% for the AMPKβ1-regulatory subunit. Each of these AMPK genes was expressed in all goldfish tissues examined (Fig. 1). AMPKα1a was expressed most highly in brain, kidney and intestine(Fig. 1A), AMPKα1b was expressed most highly in brain, kidney and gill(Fig. 1B) and AMPKβ1 was expressed at relatively constant levels across tissues with the highest expression in brain (Fig. 1C).

Goldfish remained quiescent throughout the hypoxia exposure and no fish deaths were observed. Less than 4% of all fish exposed to hypoxia showed signs of distress or lost equilibrium during the hypoxia exposure and these fish were removed from the treatment and placed into a well-aerated tank for recovery and not included in the data analysis. There were no significant effects of hypoxia exposure on blood [Hb] or Ht(Table 1). MCHC decreased significantly compared with normoxia at 1 h hypoxia exposure and remained lower than normoxic values for the duration of the hypoxia exposure(Table 1). Plasma [lactate]increased significantly by ∼6-fold over the first 0.5 h of hypoxia exposure and continued to rise, reaching values that were 11-fold higher than during normoxia at 12 h exposure to hypoxia(Table 1).

Liver [ATP] decreased significantly by ∼50% within the first 0.5 h of hypoxia exposure and remained at this lower level for the duration of the hypoxia exposure (Fig. 2A). Over the same time frame, [CrP] decreased significantly to nearly one-quarter of normoxic concentrations and was constant at this level for the remainder of the exposure (Fig. 2B). Total[Cr] was not significantly affected by hypoxia exposure, therefore free [Cr],calculated as the difference between total [Cr] and [CrP], increased significantly at 0.5 h exposure to hypoxia and remained elevated compared with normoxic controls for the 12 h hypoxia exposure(Table 2). Lactate concentrations in liver increased significantly by ∼4-fold over the first 2 h hypoxia exposure and continued to rise to values that were 7-fold higher than normoxia by 12 h (Table 2). pH_i decreased rapidly and significantly within the first 0.5 h of hypoxia exposure and remained lower than normoxic samples for the duration of the treatment (Table 2). Calculated [ADP_free] was elevated by 1 h exposure to hypoxia and remained elevated for up to 4 h after which point[ADP_free] was no longer significantly elevated compared with normoxic controls. Liver [ADP_free]/[ATP] and [AMP_free]followed similar patterns with values increasing significantly over the first 0.5 h of hypoxia exposure, remaining elevated throughout the 12 h hypoxia exposure (Table 2). Liver[AMP_free]/[ATP] increased rapidly following the onset of hypoxia,continued to increase over the first 4 h hypoxia exposure, then stabilized between 8 and 12 h at values that were between 11- and 15-fold higher than observed in normoxic controls (Fig. 2C). The Δ_fG′ of ATP hydrolysis in goldfish liver fell significantly over the first 0.5 h of hypoxia exposure and remained lower than normoxic controls for the duration of hypoxia exposure(Table 2).

Liver AMPK activity increased significantly by ∼5.5-fold over the first 0.5 h of hypoxia exposure, remained elevated until 8 h exposure to hypoxia to return to levels that were not significantly elevated compared with controls at 12 h exposure to hypoxia (Fig. 3A). Our western blot analysis using antibodies raised against human AMPKα sequence detected two immunoreactive bands of similar size(∼62 kDa) that varied in concert with each other. No other immunoreactive bands were detected on the western blots in the present study (full western blots not shown). Quantification of the darker band revealed no significant effect of hypoxia exposure on AMPKα protein expression in liver(Fig. 3B). Hypoxia exposure had no significant effect on AMPKα1a or AMPKα1b mRNA

expression; however, a significant decrease in AMPKβ1 mRNA was noted at 1, 2, 4 and 12 h hypoxia exposure (Table 3).

During hypoxia exposure, there was a rapid (by 0.5 h) and significant 2-fold increase in phospho-eEF2, which remained elevated for the first 2 h of hypoxia exposure and waned following 4 h hypoxia exposure(Fig. 4A,B). There was no significant effect of hypoxia exposure on total eEF2 quantity in goldfish liver (representative western blot shown in Fig. 4B). Protein synthesis rates in liver, as assessed by ³H-leucine incorporated into protein in cell-free extracts, decreased rapidly and significantly over the first 0.5 h of hypoxia exposure and remained depressed compared with the normoxic controls for the full duration of the hypoxia exposure(Fig. 4C).

Muscle [ATP] did not change significantly in response to hypoxia exposure(Fig. 5A). Hypoxia exposure caused a significant drop in muscle [CrP] by 2 h hypoxia, which remained lower than normoxic values throughout the hypoxia exposure(Fig. 5B). Total [Cr] levels were not affected by hypoxia exposure (data not shown) thus calculated free[Cr] increased during hypoxia exposure(Table 4). Muscle [lactate]increased significantly by ∼4-fold within the first 1 h of hypoxia exposure and continued to increase throughout the hypoxia exposure(Table 4). Muscle pH_i fell significantly by 1 h hypoxia exposure and remained lower than normoxic controls for the duration of treatment(Table 4). There were no significant effects of hypoxia exposure on calculated muscle[ADP_free], [ADP_free]/[ATP], or[AMP_free]/[ATP] (Table 4; Fig. 5C). Muscle[AMP_free] was not significantly affected by hypoxia exposure until 12 h, where concentrations increased significantly by nearly ∼5.5-fold(Table 4). Furthermore, there was no significant effect of hypoxia exposure on theΔ _fG′ of ATP hydrolysis in goldfish muscle until 12 h hypoxia exposure, where Δ_fG′ decreased significantly relative to the normoxic controls (Table 4).

Unlike the response observed in liver, AMPK activity in muscle was not affected significantly by 12 h of hypoxia exposure(Fig. 6A). Only a single immunoreactive band at ∼62 kDa was observed in muscle (see representative western blot analysis, Fig. 6C)and hypoxia exposure had no significant effect on the amount of AMPK protein(Fig. 6B,C).

There was no significant effect of 0.5, 8 or 12 h hypoxia exposure on AMPK activity in brain, gills or heart (Table 5).

In order to survive hypoxia exposure, hypoxia-tolerant organisms must reorganize cellular metabolism to limit ATP demands to match the limited capacity for O₂-independent ATP production. It has been speculated(Bartrons et al., 2004; Bickler and Buck, 2007; Rider et al., 2006) but never experimentally determined that AMPK may play an important role in co-ordinating the hypoxic cellular response in hypoxia-tolerant animals. The present study is the first to show that AMPK activity increases in response to short-term, severe hypoxia exposure in a hypoxia-tolerant animal, and that this activation is associated with a reduction in protein synthesis rates,potentially mediated through the phosphorylation of eEF2. There was a close temporal relationship between calculated increases in[AMP_free]/[ATP] and the activation of AMPK in goldfish liver(Table 2; Fig. 3), supporting the notion that a disruption of cellular energy status is essential for activation of AMPK. Responses of AMPK to short-term, severe hypoxia exposure were tissue-specific, with responses observed only in the liver and not in other tissues.

It has been suggested that maintenance of a stable cellular [ATP] during hypoxia exposure is the hallmark measure of a hypoxia-tolerant animal(Boutilier, 2001; Hochachka et al., 1996);however, this may be an over simplification. In goldfish, hypoxia exposure caused liver [ATP] to decrease by nearly half during the first 0.5 h but following this initial drop, liver [ATP] stabilized for the duration of hypoxia exposure (Fig. 2). These results are in general agreement with the results of Busk and Boutilier who showed in isolated eel hepatocytes that anoxia caused an initial decrease in [ATP] followed by a stabilization at a new, lower level(Busk and Boutilier, 2005). By contrast, Krumschnabel et al., demonstrated that exposure of isolated goldfish hepatocytes to anoxia did not result in a decrease in [ATP], whereas the same preparation exposed to chemical anoxia (sodium cyanide), showed a decrease in[ATP] (Krumschnabel et al.,1997). This latter decrease in [ATP] was modest when compared with the large decreases in [ATP] observed in anoxia-exposed hepatocytes isolated from the hypoxia-intolerant rainbow trout (Oncorhynchus mykiss). The fact that [ATP] is maintained in the liver after an initial disruption, rather than falling to fatally low concentrations highlights the ability of the hypoxia-tolerant goldfish to enter a state of lower ATP turnover. Furthermore,the decrease in [ATP] in fish liver, but not muscle(Fig. 5A), during hypoxia has been described previously in goldfish (van den Thillart et al., 1980) and sole, Solea solea(Dalla Via et al., 1994). These authors postulated that the low [CrP] observed in liver results in an inability to adequately buffer [ATP] during the onset of hypoxia. This hypothesis agrees with the present study where liver had ∼3-fold lower[CrP] than muscle (cf. Fig. 2Band Fig. 5B), and [CrP]decreased in goldfish liver over the same time period as the decline in [ATP](Fig. 2). Overall, the initial decline in liver [ATP] results from its hydrolysis in the face of a blunted capacity for ATP production, which leads to the observed accumulation of its breakdown products, [ADP_free] and [AMP_free](Table 2). In agreement,[ADP_free]/[ATP] and [AMP_free]/[ATP] increased significantly during the first 0.5 h and remained elevated for the duration of the hypoxia exposure (Fig. 2; Table 2).

Hypoxia exposure caused a significant disruption of cellular phosphorylation potential in goldfish liver with a rapid (within 0.5 h) and dramatic decrease in Δ_fG′ of ATP hydrolysis occurring upon hypoxia exposure (Table 2). Following the initial decrease, there was a stabilization ofΔ _fG′ of ATP hydrolysis at approximately –54 kJ mol^–1, which is above the values calculated as being required for the function of ATPases in rat myocardium [–49, –51, –53 and –45 to –50 kJ mol^–1 for sarcolemmal Na⁺/K⁺-ATPases, Ca²⁺-ATPases, sarcoplasmic reticulum Ca²⁺-ATPases and actomyosin-ATPases, respectively(Kammermeier, 1987; Kammermeier, 1993)]. Species differences in ΔfG′ of ATP hydrolysis requirements exist(Pörtner et al., 1996)and the actual requirements of goldfish ATPases are not known; however, our data suggest that the free energy of ATP hydrolysis in goldfish liver during hypoxia exposure is maintained at a balanced level that continues to allow for the function of integral cellular processes, albeit probably at substantially reduced levels.

Clearly, the liver is capable of readjusting metabolism after a short period of transition and it has been proposed that this occurs through the co-ordinated depression of ATP hydrolysis and increased glycolytic flux to support ATP production (Hochachka et al.,1996). The activation of AMPK, as observed in the present study(Fig. 3A), has been hypothesized to co-ordinate these events(Hardie, 2007). AMPK activation decreases rates of cellular anabolism, in particular the rates of protein synthesis, and upregulates PFK activity in rat cardiomyocytes(Marsin et al., 2000) and increases GLUT-4 transcription and insertion into rat skeletal muscle membranes (Holmes et al.,1999; Kurth-Kraczek et al.,1999). Overall, these effects on PFK and GLUT transporters, if they occur in fish, should enhanced glycolytic ATP production. Furthermore,PFK is allosterically activated by increasing [AMP_free] and[ADP_free] and inhibited by high [ATP], possibly further enhancing glycolysis and lactate production during hypoxia exposure.

Lactate accumulation in goldfish liver during our hypoxia exposure occurred at a relatively slow rate with a significant 4-fold increase occurring at only 2 h (Table 2). This slow accumulation of lactate tends to argue against a substantial increase in liver glycolytic flux; however, it must be noted that our experimental design does not allow us to assess liver glycolytic flux during hypoxia exposure in goldfish. To calculate glycolytic flux we would need to know the rates of lactate uptake or release from the liver, which is difficult to assess in vivo. The observed plasma to liver [lactate] gradient (estimated liver intracellular lactate is 7.2 mmol l^–1 after accounting for an intracellular water content of 0.8 ml g^–1 wet tissue cf. 10.4 mmol l^–1 in plasma) (Tables 1 and 2), opens the possibility that measured liver lactate is of plasma origin and not endogenously produced,although a number of factors play into dictating lactate movement across membranes (Wang et al., 1996). In addition, we exposed the goldfish to severe hypoxia, not anoxia, therefore O₂ is still available for mitochondrial respiration and ATP production, although possibly occurring at reduced levels. Overall,measurements of liver O₂ consumption and liver lactate production rates would be useful to determine the relative roles of metabolic rate depression, enhanced glycolysis and oxidative phosphorylation in longer-term(>1 h) hypoxia survival. However, it remains reasonable to suggest that prolonged hypoxia-survival necessitates a decrease in ATP utilization that can be supplied by either moderately enhanced glycolytic flux or sustained oxidative phosphorylation and that these events appear to be co-ordinated by the activation of AMPK.

The mechanism for the observed AMPK activation in goldfish liver is likely to be post-translational modification through phosphorylation of theα-subunit at Thr-172 (Beauloye et al., 2001; Rider et al.,2006). Decreases in liver [ATP] may be a requisite for the observed AMPK activation as ATP binds to the same allosteric domain as AMP(Hardie, 2007; Scott et al., 2004)competitively inhibiting AMPK (Corton et al., 1995; Hardie et al.,2006). The concurrent increase in [AMP_free] and the decrease in [ATP] observed in goldfish liver may thus be required to activate AMPK. To ascertain whether the increase in AMPK activity shown in Fig. 3A was due to phosphorylation, we screened several anti-phospho-Thr-172 AMPK antibodies but discovered that the antibodies were unable to detect the phosphorylated protein in any goldfish tissue tested. Regardless, the rapid and large-scale changes in activation state of liver AMPK are probably only possible by post-translational modification. No changes in total AMPKα protein or mRNA expression and decreases in AMPKβ1 mRNA levels were observed during the 12 h hypoxia exposure (Table 3; Fig. 3B)therefore, upregulation of protein expression cannot explain AMPK activation. The lack of change in AMPK α-subunit expression concurs with previously published findings from human glioblastoma cells, which show no change in AMPKα1 mRNA or protein expression in response to hypoxia exposure and only demonstrate an upregulation of AMPKα2 isoform protein and mRNA after prolonged hypoxia exposure [>24 h(Neurath et al., 2006)].

Regardless of expression pattern, it is known that the activation of AMPK decreases cellular protein synthesis rates, in part through direct phosphoryation of eukaryotic elongation factor-2 kinase [eEF2K(Browne et al., 2004; Horman et al., 2002)], which in turn phosphorylates eEF2 and renders eEF2 unable to bind ribosomes. In the present study, we demonstrate a rapid (within 0.5 h) and significant increase in phosphorylation of eEF2 at Thr-56 in livers of hypoxic goldfish(Fig. 4A). This increase in phosphorylation is temporally associated with a significant decline in the rate of ³H-leucine incorporation into new proteins in cell-free extracts, which fell to ∼70% of normoxic values by 0.5 h hypoxia exposure and continued to fall to ∼93% of normoxia by 4 h hypoxia exposure(Fig. 4B). Protein synthesis,for its part, accounts for 20–30% of total ATP-coupled O₂demand (Bickler and Buck, 2007)and has been shown to decrease by ∼90% in anoxia-tolerant hepatocyte cultures (Land et al., 1993)and by 56–95% in the liver of crucian carp(Smith et al., 1996) and Amazonian cichlids [Astronotus ocellatus(Lewis et al., 2007)] during anoxia/hypoxia exposure. Interestingly, in the present study, both the increase in phosphorylation of eEF2 and the decline in ³H-leucine incorporation into proteins occurs over the same timescale as the increase in AMPK activity (Fig. 3)suggesting that the hypoxia-induced reductions in protein synthesis may be mediated by the activation of AMPK. To demonstrate a causal relationship between AMPK activation and inhibition of protein synthesis in hypoxic goldfish, direct manipulation of AMPK activity using the pharmacological activator 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR) must be performed.

The large decrease in liver pH_i measured in goldfish during hypoxia exposure (Table 2) may also contribute to the regulation of protein synthesis and metabolic rate depression. Tissue acidosis has been shown to cause an increase in eEF2K activity and eEF2 phosporylation, reducing protein synthesis rates(Dorovkov et al., 2002). In addition, beyond the specific effects of acidosis on protein synthesis,decreases in pH_i and extracellular pH are thought to contribute to initiating and sustaining metabolic rate depression in vertebrates and invertebrates through general effects on enzyme function or membrane transport(e.g. Reipschläger and Pörtner,1996), although not all studies support this notion (e.g. Brooks and Storey, 1989).

Because it appears that AMPK activation can affect metabolically costly processes, like protein synthesis, in hypoxia-tolerant fish and facilitate metabolic rate depression, it is of interest to consider other proteins or pathways, which are important in metabolic rate depression and may be activated by AMPK. For instance, second to protein synthesis, iono-regulation is the largest energy sink in the cell comprising ∼20% of ATP demand in hypoxia-tolerant hepatocytes of the western painted turtle, Chrysemys picta belli (Hochachka et al.,1996). In goldfish hepatocytes, Na⁺ pump activity and K⁺ leak pathways are downregulated in a co-ordinated manner during chemical anoxia for energy conservation purposes(Krumschnabel et al., 1996). This ability of hypoxia-tolerant cells to manipulate ion regulatory processes contributes to a large degree to metabolic rate depression and represents an appealing target for regulation by AMPK. Interestingly, epithelial Na⁺ channel currents in Xenopus oocytes and collecting duct cells in mice are inhibited in an AMPK-dependant manner(Carattino et al., 2005)demonstrating that some iono-regulatory action of AMPK is known.

AMPK α and β subunits are expressed in all tissues examined(Fig. 1) with the highest levels of mRNA being present in the brain, kidney, intestine and gill. However, this tissue-specific expression pattern does not translate into detectable differences in AMPK activation in these tissues during short-term O₂-deprivation. Unlike results demonstrated in the goldfish liver,no activation of AMPK was observed in muscle, brain, heart or gill during 12 h of severe hypoxia exposure (Fig. 6; Table 5). These results are in contrast to those obtained in hypoxia-sensitive mammalian models (Kudo et al., 1995; McCullough et al., 2005; Mu et al., 2001) where AMPK activation in muscle, brain and heart was observed in response to hypoxia exposure. In agreement with our results, the brain and heart of both killifish(Fundulus grandis) and trout (Salmo gairdneri) showed fewer signs of metabolic stress when exposed to hypoxia than did the skeletal muscle or liver (Dunn and Hochachka,1986; Martinez et al.,2006) indicating that not all tissues respond in a similar fashion to hypoxia exposure. There are a number of potential explanations for this tissue-specific AMPK activation in goldfish. First, our goldfish were exposed to severe hypoxia rather than complete anoxia thus differential shunting of blood to these organs during hypoxia exposure may explain why the brain, heart and gill displayed no activation of AMPK. Upon hypoxia exposure in fish,essential tissues receive increased blood flow and, therefore,O₂-delivery (Booth,1979; Gamperl et al.,1995; Soengas and Aldegunde,2002) and consequently may not experience a metabolic stress to the same degree as liver. Second, the duration of hypoxia exposure may not have been long enough to observe activation of AMPK. Third, AMPK maybe regulated in a tissue-specific fashion with either different upstream regulating kinases expressed in different tissues or the level of cellular disruption required, e.g. degree of increase in (AMP_free)/(ATP), to observe AMPK activation may differ between tissues.

Within muscle, there was no significant increase in[AMP_free]/[ATP] (Fig. 5C) and no apparent activation of AMPK during the 12 h exposure to hypoxia. The maintenance of high muscle [ATP] during hypoxia exposure, as seen in other studies (Fig. 5A)(Richards et al., 2007; van Ginneken et al., 1995; Zhou et al., 2000), may impede AMPK activation since, as mentioned previously, ATP competitively inhibits AMP binding to AMPK. Indeed, in goldfish muscle [AMP_free]/[ATP] ratios were unaltered by hypoxia exposure at all sampling times, as were[ADP_free] and [ADP_free]/[ATP] measurements(Fig. 5C; Table 4). Additionally, there was also no significant change in [AMP_free] orΔ _fG′ of ATP hydrolysis until 12 h hypoxia(Table 4), suggesting that only at >12 h hypoxia exposure might goldfish muscle experience an energy stress great enough to result in the activation of AMPK and the need to activate biochemical means of reducing ATP demands. Longer-term hypoxia exposures are needed to determine if AMPK is activated in these tissues and plays a role in hypoxic survival.

AMPK has been proposed as an appealing candidate for co-ordinating the metabolic responses of tissues to hypoxia exposure in tolerant organisms(Bartrons et al., 2004; Bickler and Buck, 2007; Rider et al., 2006). Indeed,AMPK activity increased in liver in response to hypoxia exposure and the characteristic interactions between AMPK and the downregulation of protein synthesis were in place and responded to hypoxia exposure. These responses were tissues-specific with no observed activation of AMPK in brain, gill,heart or muscle. AMPK activation was closely associated with increased[AMP_free] and decreased [ATP], suggesting that the ratio of these adenylates may have been important for activation. The decreased rates of protein synthesis, a well-known component of metabolic depression, combined with the phosphorylation of eEF2, a downstream target of AMPK, potentially implicate AMPK in the cellular effort to suppress metabolism in tolerant species exposed to hypoxia.

We gratefully acknowledge Milica Mandic and Ben Speers-Roesch for their valuable assistance during sampling. This work was funded by a Natural Science and Engineering Research Council (NSERC) of Canada Discovery Grant to J.G.R.