Crucian carp were collected in late summer from the Tjernsrud pond near Oslo, Norway, using nylon net cages. Fish were held at the aquarium facility at the Department of Biosciences, University of Oslo, in tanks with a semi-closed recirculation system supplied with dechlorinated, aerated Oslo tap water in a room with a 12:12 h light:dark cycle. The water temperature varied seasonally, but was kept constant during anoxia exposure, as described below. The fish were fed daily with commercial carp food (Tetrapond, Tetra, Melle, Germany), and acclimated to indoor conditions for at least 4 weeks prior to exposure. All experimental procedures were approved by the Norwegian Animal Health Authority (FOTS ID 16063).

All fish used in the present study were from the same experimental exposure as previously described (Johansen et al., 2023). We randomly selected crucian carp of both sex (n = 18, weight = 27 ± 8 g) that were transferred to air-tight containers, fasted and acclimated with aerated water (8°C) flowing through the tanks for 48 h prior to anoxia exposure. Anoxia was achieved by continuous low bubbling of nitrogen into the tanks, and the oxygen level monitored using a Firesting fiber-optic oxygen meter with an oxygen probe (PyroScience GmbH, Aachen, Germany). Oxygen levels were maintained at <0.1% of air saturation and the tanks kept anoxic for 5 days. The water temperature was maintained between 8.2°C and 8.6°C throughout the entire experiment. After 5 days of anoxia, nitrogen bubbling was replaced by air bubbling in the tanks, causing the water to become fully air saturated within 1 hour. We collected tissue after 5 days of anoxia; 5 days of normoxia and 5 days anoxia followed by 1 day reoxygenation. Fish were euthanized by a quick blow to the head, the spinal cord was cut, and brain and liver tissues dissected out in the mentioned order. The tissues were snap-frozen in liquid nitrogen and stored at −80°C until further analysis. Crucian carp can withstand at least 2 weeks of anoxic exposure at 8°C (Nilsson, 1990) and we did not observe any mortality during the experiment.

Frozen crucian carp brain and liver tissues from 6 fish per experimental group were homogenized with a pestle in ice-cold SILAC™ Phosphoprotein Lysis buffer (Invitrogen, Carlsbad, CA, USA). Homogenized tissue was incubated 5–10 min on ice and then at −80°C, the buffer volume was adjusted to 60 mg tissue/mL lysis buffer and the lysate cleared by centrifugation (18,000 g for 15 min). Total brain protein content was measured with a Detergent Compatible Bradford Assay Reagent (Pierce, Rockford, IL, USA) at 595 nm and total heart protein content was measured with a BCA assay (Pierce, Rockford, IL, USA) at 570 nm, both with BSA as a standard. From each biological replicate, 1 mg protein was precipitated with 5 volumes ice-cold acetone at −20°C overnight. Precipitated protein was pelleted by centrifugation (13,000 g for 15 min), the acetone was aspirated and the pellet let to air dry before resuspension in 6 M urea in 100 mM ammonium bicarbonate. Cystines present in the sample were reduced with 10 mM DTT at 30°C for 60 min and alkylated with 30 mM iodoacetamide at 22°C for 1 hour in the dark. The reaction was quenched with 30 mM DTT at 30°C for 30 min and the sample diluted with 50 mM ammonium bicarbonate before digestion with approximately 1:100 (w/w) trypsin Gold Mass Spec Grade (Promega, Madison, WI, USA) at 37°C overnight. The digest was finally quenched with 1% formic acid and the peptides cleaned by solid-phase extraction (SPE) using a Strata C18-E cartridge (55 μm, 70 Å, Phenomenex, Værlose, Denmark).

Dried tryptic peptide samples were dissolved in loading buffer (1 M glycolic acid, 6% trifluoroacetic acid, 5% glycerol, and 80% acetonitrile) under continuous shaking. TiO₂ beads (Titansphere, TiO_2, GL Sciences Inc, Japan) were washed in loading buffer 3 times before transferring them to the dissolved tryptic peptide samples. After 1 hour of continuous shaking, the supernatant was collected and transferred to a new tube containing freshly washed TiO₂ beads for a second incubation. The TiO₂ beads were collected separately and gently washed sequentially with 200 μL of loading buffer, 200 μL 80% acetonitrile/2% trifluoroacetic acid, 200 mM ammonium glutamate, and 200 μL of 50% acetonitrile/1% trifluoroacetic acid. The TiO₂ beads were dried and bound peptides were eluted sequentially in 10 min at first with 50 μL of 10% ammonium hydroxide, pH 11.7, then with 50 μL of 15% ammonium hydroxide/60% acetonitrile, and finally with 50 μL of 1% pyrrolidine. Eluted peptides were acidified by adding 75 μL 50% formic acid and cleaned up using ZipTip-C18 (Millipore, Billerica, MA, USA).

The peptide samples were analyzed with an Ultimate 3,000 nano-UHPLC system (Dionex, Sunnyvale, CA, USA) coupled to a Q Exactive mass spectrometer (ThermoElectron, Bremen, Germany) equipped with a nano electrospray ion source. Liquid chromatography (LC) separation was performed with an Acclaim PepMap 100 column (C18, 3 µm beads, 100 Å, 75 μm inner diameter, 50 cm) (Dionex, Sunnyvale CA, USA). We used a flow rate of 300 nL/min with solvent A (0.1% formic acid) and solvent B (0.1% formic acid/90% acetonitrile) and applied a solvent gradient of 4%–35% of solvent B in 160 min, to 50% solvent B in 20 min and finally to 80% of solvent B in 2 min. We operated the mass spectrometer in data-dependent mode to switch automatically between MS and MS/MS acquisition. Survey full scan MS spectra (from m/z 300–2000) were acquired with the resolution R = 70,000 at m/z 200, after accumulation to a target of 1e6. The maximum allowed ion accumulation times were set to 100 m. The used method allowed sequential isolation of up to the 10 most intense ions, depending on signal intensity (intensity threshold 1.7e4), for fragmentation using higher-energy collision induced dissociation (HCD). The target value was 10,000 charges and the resolution set to R = 17,500 with NCE 28. Target ions that were already selected for MS/MS were dynamically excluded for 60 s and the isolation window was m/z = 2 without offset. The maximum allowed ion accumulation for the MS/MS spectrum was 60 m. We enabled the lock mass option in MS mode for internal recalibration during the analysis and accurate mass measurements.

All MS/MS samples were searched against the Carassius auratus database (NCBI; all taxons; 96,703 entries), with PEAKS X+ software version 10.5 (Bioinformatics Solutions, Waterloo, ON, Canada). The following parameters were used: digestion enzyme, trypsin; maximum missed cleavage, 1; fragment ion mass error tolerance, 0.1 Da; parent ion error tolerance, 10.0 ppm; and fixed modifications, carbamidomethylation. Oxidation of methionine and acetylation of the N-terminus were specified as variable modifications and the maximum number of PTMs was set to 2. A false-discovery rate of 1% was applied to the datasets. Samples that contained less than 4,200 (brain) or 2000 (liver) peptides were removed, resulting in five (brain) and four (liver) samples from each treatment group that were used for label-free quantitation (LFQ). For LFQ, normalization by total ion count (TIC) was applied, and the following parameters used for filtration: peptide quality ≥2; average area ≥ 1e5; charge between 2 and 5; used peptides ≥1. Phosphorylated peptides passing the applied filters were imported to the Perseus software (Tyanova et al., 2016) for statistical analysis. Peptides had to be identified in at least three replicates in at least one experimental group to be considered (Tyanova and Cox, 2018). Any missing values were replaced by 1e4 before log2 transformation. The overall expression patterns were evaluated by principal component analysis (PCA) using the web tool ClustVis (Metsalu and Vilo, 2015). Data were centered and scaled by unit variance scaling, and singular value decomposition (SVD) with imputation applied to calculate the principal components. Regulation of phosphorylation was assessed on the peptide level by one-way ANOVA and multiple testing corrected for with Tukey’s HSD test applying FDR 5% in Perseus. Phosphopeptides varying by a fold change ≥2 were kept. Hiearchichal clustering was conducted with ClustVis (Metsalu and Vilo, 2015) on normalized and ln(x) transformed data. Rows were centered and scaled by unit variance before clustering using correlation distance and average linkage. Venn diagrams were created using BioVenn (Hulsen et al., 2008). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifier PXD033061.

Goldfish protein sequences of the significantly regulated phosphoproteins were uploaded to BlastKOALA (Kanehisa et al., 2016) to acquire the Uniprot gene names. The entry names were then matched with the corresponding zebrafish names for enrichment analysis. g:Profiler (version e101_e.g.,48_p14_baf17f0) was used for functional profiling with g:SCS multiple testing correction method and a significance threshold of 0.05 (Raudvere et al., 2019). The zebrafish entry list was imported as an unordered query including the entire ANOVA protein list. The KEGG pathway database (Kanehisa, 2019) was used to visualize the relevant enriched pathways.

To explore the impact of oxygen deprivation on the phosphoproteome of crucian carp, we exposed wild-caught crucian carp to 5 days normoxia (air-saturated water), 5 days anoxia (<0.1% of air saturation) and 5 days anoxia followed by 1 day of reoxygenation. Phosphorylated peptides of brain and liver tissues were enriched using TiO₂ beads and analyzed by LC-MS/MS for label-free quantitation (Supplementary Tables S1, S2). Only phosphopeptides detected in at least three replicates within at least one experimental group were used for further statistical analysis and revealed 4,200 phosphopeptides in the brain and 1,287 phosphopeptides in the liver. Unsupervised PCA (on log2 transformed data) was applied to first examine the global effect on the phosphoproteomes of brain and liver. Interestingly, there was no apparent clustering of experimental groups in the brain phosphoproteome indicating few anoxia-specific changes (Figure 1A). In contrast, the liver phosphoproteome separated into the experimental groups clearly indicating strong effect with anoxia-reoxygenation (Figure 1B).

Next, we applied statistical analysis in combination with filtering out peptides regulated by fold change <2 to focus on the peptides that changed the most in response to anoxia-reoxygenation. The statistical analysis was performed on peptide level rather than protein level since the latter strategy may occult changes of individual peptides within the same protein. Although the total amount of detected phosphopeptides was considerably higher in the brain (4,200 peptides) than in the liver (1,287 peptides), the number of regulated phosphopeptides was much higher in the liver. We found that 109 phosphopeptides in the brain and 395 phosphopeptides in the liver were significantly changed (FDR <5%) compared to normoxia (Figure 2; Supplementary Tables S3, S4). In the brain, 68 phosphopeptides changed in anoxia compared to normoxia (40 increased and 28 decreased in phosphorylation level), and 59 peptides changed in phosphorylation levels in reoxygenation compared to normoxia (49 increased and 10 decreased). The number of phosphopeptides changing in phosphorylation level in anoxia compared to normoxia was 286 in the liver (194 increased and 92 decreased), and 215 liver peptides changed in phosphorylation levels in reoxygenation compared to normoxia (172 increased and 42 decreased). Hierarchically clustering of the significant peptides Figure 2 showed most grouping of peptides according to experimental groups in the liver (Figure 2B).

Comparison of the changed phosphopeptides revealed that few phosphoproteins were changed at the same phosphosite in both tissues in response to anoxia or reoxygenation (Figure 3). Only 12 phosphopeptides corresponding to 11 phosphoproteins showed similar regulation in both brain and liver (Table 1). Among them, we found proteins involved in kinase signaling (AMPK and MAPK3), apoptotic control (CAST and BADB) and translation initiation (LARP1 and EIF4EBP2), suggesting that these processes are a common response in both tissues to anoxia-reoxygenation conditions in the crucian carp (Table 1).

Differentially changed phosphoproteins (FDR <0.05, fold change >2), including both increased and decreased phosphoproteins, were uploaded to the g:Profiler software for functional enrichment analysis. In response to anoxia-reoxygenation in the brain, enriched gene ontology (GO) terms included L1CAM interactions (Reactome), central nervous system neuron development (biological process) and GTPase binding (molecular function) (Supplementary Table S5). In the liver, the insulin signaling pathway (KEGG pathway) was among the most significantly enriched pathway in response to anoxia-reoxygenation (Supplementary Table S6; Supplementary Figure S1). PKA activation in glucagon signaling (Reactome) and AMPK activity (molecular function) were also enriched in the liver during anoxia-reoxygenation (Supplementary Table S6).

Phosphopeptides belonging to AMPK were among the most changing peptides in terms of fold change in both brain and liver (Table 3; Supplementary Tables S3, S4), and AMPK signaling was an enriched GO term in the liver (Supplementary Table S6). Consequently, we investigated in detail the phosphoproteins implicated in the closely related pathways of insulin signaling, PKA activation and AMPK signaling. Because glucose flux is sensitive to insulin and glucagon, and both signaling pathways were enriched in the liver phosphoproteome, we also examined the regulation of glycolytic enzymes in more detail.

AMPK is a well-known key regulator of metabolic depression (Hardie and Ashford, 2014) and was differentially phosphorylated in this study. The observed increase in AMPK phosphorylation in anoxic brain and liver stimulates AMPK activity, and are in line with previously observed AMPK alpha phosphorylation of the activating Thr-172 in anoxic crucian carp brain, heart and liver (Stensløkken et al., 2008) and hypoxic goldfish liver (Jibb and Richards, 2008). AMPK phosphorylation was still high after 1 day of reoxygenation in brain and liver (Figure 4), consistent with the maintained high phosphorylation levels of AMPK in crucian carp brain after 7 days of reoxygenation (Stensløkken et al., 2008). The phosphorylation status of AMPK suggests maintained activity throughout reoxygenation. However, we have previously found that within 3 hours of reoxygenation the hepatic ATP level even overshoots the normoxic level and that the AMP level has fully recovered (Dahl et al., 2021). The resultant high ATP:AMP ratio should mean that ATP outcompetes AMP for binding to AMPK, inactivating the enzyme regardless of phosphorylation status (Li et al., 2017). Taken together with the phosphorylation pattern of downstream AMPK targets during reoxygenation (Figure 4), this suggests that AMPK, even if still phosphorylated, is not active during reoxygenation in the crucian carp.

mTORC1 has recently gained attention due to its key role in metabolic stress responses in hypoxia-tolerant animals (Wu and Storey, 2021). In anoxia-intolerant mammals, oxygen depletion triggers AMPK activation, which in turn phosphorylates the TSC1/2 complex and inhibits mTORC1 activity (Wullschleger et al., 2006). In contrast, hypoxia-tolerant naked mole rats (Heterocephalus glaber) and red-eared slider turtles (Trachemys scripta elegans) experience increasing mTORC1 activities in response to hypoxia and anoxia, respectively (Szereszewski and Storey, 2018; Wu and Storey, 2021). In the case of crucian carp, AMPK is activated only in response to anoxia, and not hypoxia (Stensløkken et al., 2008). Even though we did not detect mTORC1 phosphorylation in this present study, our findings of reduced phosphorylation of mTORC1-associated proteins (PRAS40, TSC1) and downstream targets (EIF4E-BPs, LARP1) collectively suggest that mTORC1 activity in the crucian carp liver is reduced (Figure 4). It seems likely that this inhibition is a result of AMPK activation.

Protein synthesis in crucian carp has been shown to be relatively unaffected but possibly slightly suppressed in brain, but heavily inhibited (>95%) in liver during anoxia (Smith et al., 1996). Our results suggest that this stall in translation could be regulated through phosphorylation of translation initiation binding proteins (Table 1). The majority of the identified EIF4E-BP phosphopeptides returned to normoxic phosphorylation levels after 1 day of reoxygenation (Table 3) in line with the reported normoxic translational rates in crucian carp liver after 24 hours reoxygenation (Smith et al., 1996). In addition, dephosphorylation of LARP1 has been attributed to inhibiting translation (Fonseca et al., 2015), suggesting that the observed reduction in LARP1 phosphorylation in anoxic crucian carp liver (Figure 4A) also participates in suppressing translation. Our data therefore suggest that brain protein synthesis is reduced to some extent, since EIF4E-BP and LARP1 phosphorylation decreased in the anoxic brain (Table 2).

Glucose supplied by from a huge hepatic glycogen store is the main energy source for most crucian carp tissues during anoxia and crucial for long-term survival (Vornanen and Haverinen, 2016). In response to anoxia-reoxygenation, we found changes in levels of phosphoproteins involved in hepatic glycogen metabolism (PHKA2, PGM1) and glucose flux in glycolysis/gluconeogenesis (PFKFB1, GPAT3). In the liver, phosphorylation of PHKA2 (alpha subunit of phosphorylase b kinase) increased during anoxia (Table 4). Interestingly, phosphorylation of PHKB (beta subunit of phosphorylase b kinase) at Ser-684 increased in the anoxic brain (Supplementary Table S3), in contrast to the reduced phosphorylation levels at Ser-701 in the liver (Table 4). The fact that this was the only glucose-related enzyme that appeared to be regulated by phosphorylation in the brain is interesting and probably reflects that crucian carp brain is well protected from energy depletion during anoxia.

The observed increase in phosphorylation of PFKFB-1 in the liver indicates a possible degradation of fructose-2,6-bisphosphate during anoxia and reoxygenation, and is probably a result of glucagon-mediated PKA phosphorylation (Kurland and Pilkis, 1995; Janah et al., 2019). Fructose-2,6-bisphosphate has been shown to be the most important allosteric factor for PFK1 activity (Uyeda et al., 1981), and reduced concentrations are coupled to attenuated glycolysis and stimulated gluconeogenesis (Uyeda et al., 1981; Ros and Schulze, 2013). While this makes sense during reoxygenation, when liver tissue starts to rebuild the glycogen stores (Vornanen and Haverinen, 2016), it was more surprising to find the same PFKFB-1 phosphorylation pattern also in anoxia. However, reduced concentrations of fructose-2,6-bisphosphate have previously been observed in anoxic goldfish liver (Storey, 1987), supporting degradation of this allosteric factor. Taken together, we hypothesize that simultaneous glycolysis and gluconeogenesis is occurring in hepatic tissue during anoxia-reoxygenation (Moreno et al., 2014).

One of the proteins that exhibited high fold changes in both brain and liver during anoxia exposure and recovery was the BCL2-associated agonist of cell death (BADB). Phosphorylation at Ser-118 of BADB increased during anoxia and reoxygenation in both tissues (Table 1) and has been reported to subsequently inactivate the protein by binding to 14-3-3 proteins, thereby hindering apoptosis and stimulating cell survival (Downward, 1999; Datta et al., 2000). This finding is in line with the increased gene expression of the anti-apoptotic proteins BCL-2 and BCL-XL during anoxia in the anoxia-tolerant turtle (Krivoruchko and Storey, 2015). Furthermore, BCL-2 knockdown in turtle neuronally enriched primary cell cultures caused increased apoptosis and ROS levels during anoxia-reoxygenation (Kesaraju et al., 2014). A moderate increase in cell death after reoxygenation has been observed in crucian carp brain, but not coinciding with changes in caspase-3 transcript expression (Lefevre et al., 2017). Part of the caspase-3 activation occurs through BCL-2-mediated cytochrome c release and downstream activation of caspase-3 (Birkinshaw and Czabotar, 2017). Similarly as BCL-2 suppression in turtle activates caspase-3 (Kesaraju et al., 2014), phosphorylation and inactivation of BADB would contribute to the inhibition of this pathway, and could explain why the increased neuronal cell death is probably not linked to caspase-3. Other proteins involved in apoptotic control did also change in phosphorylation levels during anoxia-reoxygenation in the brain and the liver. For example, apoptotic chromatin condensation inducer in the nucleus (ACIN1) exhibited one of the most elevated phosphorylation levels in the anoxic liver compared to normoxia (Supplementary Table S4). Furthermore, programmed cell death protein 4 (PDCD4A), calpastatin (CAST), translationally-controlled tumor protein homolog (TPT1), forkhead box protein O1-A (FOXO1A) and 14-3-3- gamma protein (YWHAH) all underwent changing phosphorylation patterns in the liver (Supplementary Table S4), while PDCD4B and CAST phosphorylation were also regulated in the brain (Supplementary Table S3). Although the precise functions of these phosphorylation sites are scarcely described in the literature, they still point towards broad regulation of apoptosis in the brain and liver during anoxia and reoxygenation.

Interestingly, among the phosphopeptides that exhibited the highest fold changes in the liver during anoxia-reoxygenation we found the ROS scavenging enzymes superoxide dismutase [Cu-Zn] 1 (SOD1) and peroxiredoxin (PRDX; Supplementary Table S4). However, no functionality has so far been reported for the regulated phosphosites in either SOD1 or PRDX. In the goldfish, total SOD activity (including both cytoplasmic SOD1 and mitochondrial SOD2) was unchanged in the brain and liver during anoxia and reoxygenation (Lushchak et al., 2001). In the moderately anoxia-tolerant salamander, SOD activity did remarkably decrease during reoxygenation (Issartel et al., 2009), while antioxidant activities decreased during anoxia in anoxia-tolerant turtles (Willmore and Storey, 1997). Clearly more than one strategy for ROS handling during oxygen depletion and recovery exist, and the exact mechanisms occurring in the crucian carp remain to be elucidated.

Although a substantial higher number of phosphopeptides were identified in the brain phosphoproteome compared to the liver phosphoproteome, the majority of differentially changed phosphopeptides were found in the liver. Some of the phosphopeptides showing the biggest change in brain belonged to proteins involved in neuronal activity at the synaptic cleft. Although no functionality could be assessed to the phosphosites in this study, their differentiation still indicates a role in synaptic activity that calls for further studies. In the liver, changed phosphoproteins were related to glycolytic flux and glycogenolysis, Common in both tissues, regulation towards metabolic depression through AMPK-signaling was apparent. Moreover, our data show strong regulation of phosphosites in proteins involved in cell death regulation and ROS handling and indicate a direction towards cell survival rather than apoptosis in response to oxygen depletion and recovery. This could presumably be made possible through inhibition of apoptotic pathways and low ROS accumulation, either due to low ROS generation or high antioxidant activity. More focus should be given to the functional impact of these processes during anoxia and reoxygenation. This work also emphasizes the need for including the understudied and important liver tissue more systematically in future studies to elucidate how the crucian carp survive anoxia-reoxygenation.