Fish Physiology and Biochemistry

, Volume 39, Issue 4, pp 871–879

Slaughter of Atlantic salmon (Salmo salar L.) in the presence of carbon monoxide


    • Institute of Marine Research
    • Department of Molecular BiologyUniversity of Bergen
  • Bjørn Olav Kvamme
    • Institute of Marine Research
  • Arnt J. Raae
    • Department of Molecular BiologyUniversity of Bergen
  • Bjorn Roth
    • Department of Processing TechnologyNofima AS
  • Erik Slinde
    • Institute of Marine Research
    • Department of Chemistry, Biotechnology and Food ScienceNorwegian University of Life Science

DOI: 10.1007/s10695-012-9747-5

Cite this article as:
Bjørlykke, G.A., Kvamme, B.O., Raae, A.J. et al. Fish Physiol Biochem (2013) 39: 871. doi:10.1007/s10695-012-9747-5


The different stunning methods for Atlantic salmon can still be improved with regard to animal welfare. Salmon exposed to carbon monoxide expressed no aversive reactions towards CO as such. CO exposed fish showed an earlier onset of rigour mortis and a faster decrease in muscle pH due to depletion of oxygen during the treatment. Exposure to CO did increase the level of cortisol compared to undisturbed control fish, but the increase was less than in the water only control group. Neuroglobin, a CO binding globin, was found in salmon brain and Saccus vasculosus, a richly vascularized sac connected to the fish brain. Binding of CO to neuroglobin during sedation might possibly improve animal welfare.


Animal welfareAtlantic salmonCarbon monoxideNeuroglobinSaccus vasculosusSlaughter


In Norway, a variety of stunning methods are used on Atlantic salmon including percussive stunning (Roth et al. 2002, 2006), electricity (Roth et al. 2003), and CO2 (Erikson et al. 2006; Roth et al. 2006) either with or without live chilling. The current methods for slaughter of Atlantic salmon have recently been reviewed by European Food Safety Authorities (EFSA 2009) and Mejdell et al. (2010). The conclusion was that the current method of using CO2, with or without chilling, results in the poorest welfare for salmonids. The Norwegian ban on carbon dioxide (CO2) use will be effective from 1 July 2012.

Electricity and percussion have been extensively studied on the basis of technological and quality parameters and have been highly improved during the latter years. However, both methods have disadvantages. Electrical stimulation is known to induce injuries to the spinal cord, aorta and veins, and consequently reduces the quality of fillets (Roth et al. 2003). With percussive stunning, a major challenge is to hit the skull correctly, and problems may arise, for example, when fish are active or of different sizes. Manual percussion may be necessary when the automated procedure fails, and hitting the fish twice raises welfare issues (EFSA 2009).

Sedation or stunning of fish using gas has several advantages including improvement of the success rate of electrical or percussive stunning when used prior to these techniques. As an alternative to CO2, the use of carbon monoxide (CO) is an interesting possibility. The main toxic effect of exposure to exogenous CO results from its combination with haemoglobin (Hb) to form carboxyhaemoglobin, hence preventing the carriage of oxygen by red blood cells. The affinity of Hb for CO is more than 200 times greater than for oxygen, and at atmospheric pressure and oxygen saturation, carboxyhaemoglobin is very stable. CO also binds to other globins such as myoglobin (Mb) and neuroglobin (Ngb) (Brunori and Vallone 2007).

CO has for a long time been used in animal euthanasia. Advantages of using CO as an euthanizing agent for animals include rapid induction and painless death, no awareness of the agent and no aversive reactions as well as little or no fear in the animal (Blackmore 1993). However, the tolerance of fish to hypoxia differs markedly to that of mammals. Compared to terrestrial animals, fish are particularly hard to kill due to a general adaptation to a hypoxic environment (EFSA 2009). This is in part due to low rate of energy demand being exothermic, but also to the fact that these animals have the capability for anaerobic energy metabolism in the brain (Soengas and Aldegunde 2002). From this perspective, the Saccus vasculosus (SV) and Ngb become important. The SV is a specialized ependymovascular diverticle of the caudal hypothalamus of elasmobranchs in most bony fishes. The organ is well vascularized, and several putative functions including pressure reception and regulation, chemoreception, glucose loading, ionic regulation of the cerebrospinal fluid as well as storage and transport function have been hypothesized (Yanez et al. 1997; Sanson 1998). Ngb is a recently discovered globin mainly expressed in nervous tissue (Burmester et al. 2000; Burmester and Hankeln 2009). Ngb is also expressed in some endocrine tissues (Reuss et al. 2002) as well as in the gills of fish (Fuchs et al. 2004). The biological function of Ngb is still unclear, but it has been hypothesized that it is involved in transport and storage of O2, in scavenging of ROS (reactive oxygen species) and NO (nitric oxide), and in G-protein signalling and binding of CO (Burmester and Hankeln 2009). It has been shown that Ngb plays a protective role during hypoxia (Sun et al. 2001; Liu et al. 2009). In zebrafish and hypoxia tolerant turtles, the expression of Ngb is up regulated in response to hypoxic conditions (Milton et al. 2006; Roesner et al. 2008). In goldfish, Ngb is not affected by hypoxia, but the level of Ngb in the goldfish brain is five times that in the zebrafish brain (Roesner et al. 2008). Hence, although the biological functions of Ngb are not yet clear, a role in oxygen metabolism and hypoxia tolerance is not unlikely. We believe that Ngb may play an important role in survival in stressful situations and hypoxic conditions.

At present, there is very little information on the effects of exposure of fish to exogenous CO. However, a recently published paper (Bjørlykke et al. 2011) presents information on the effects of carbon monoxide on Atlantic salmon, regarding stress and carcase quality parameters. The paper showed that CO had a positive effect on salmon filet colour and resulted in an earlier onset of rigour and a faster decrease in pH due to lactate production. The results indicated that treatment with CO led to rapid depletion of tissue O2 concentrations in salmon. Research on the effect of CO anaesthesia of tilapia, and on post-mortem appearance of tilapia fillets from animals killed with CO, showed a significant increase in redness (a*) and lightness (L*) in fillets compared to controls (Mantilla et al. 2008). Furthermore, several other studies show that carbon monoxide enhances colour and quality of the fish (Gee and Brown 1981; Chow et al. 1998; Hsieh et al. 1998).

The objective of the present work was to study the effect of CO on Atlantic salmon in a slaughter relevant setting and determine the amount of Ngb in the brain and S. vasculosus of Atlantic salmon.

Materials and methods

Experimental set-up

Bjørlykke et al. (2011) subjected Atlantic salmon (Salmo salar L.) to CO that was diffused into the water for 20 min. The treatment of the fish was gentle, and the relative CO concentration was low (CO-L), with a gradual increase from zero to a higher level at the end of the trial (20 or 30 min). It was evident from the results shown in Bjørlykke et al. (2011) that the experimental procedure involved submitted the fish to a level of stress similar to that of handling. Hence, a further experiment has been carried out with emphasis on reducing stress from the experimental procedures. Four lidded 1,000 L square tanks (138 cm length, 108 cm width) placed indoor in a well-ventilated room were used. During acclimatization, these contained approximately 350 L of water with a water flow of approximately 15 L/min. CO water was created in outdoor feeder tanks containing 1,000 L seawater. The feeder tanks were flushed with 100 % food grade CO (Yara Praxair, Oslo, Norway) using two circular ceramic diffusers with diffuser area 15.6 cm2 (Wedge lock base unit, Point Four Systems Inc., Richmond, Canada) operated at 3 bar, placed at the bottom of the tank. Two concentrations of CO were created by diffusing CO into water for either 60 min [defined as medium CO concentration (CO-M)] or 120 min [defined as high CO concentration (CO-H)]. These solutions were used immediately. All trials were performed using approximately 35 ‰ saltwater at a temperature of 7.4 ± 0.2 °C. Controls and treatments were conducted in duplicate with four salmon for each replicate. Groups of 4 Atlantic salmon (0.8 ± 0.1 kg) were netted from holding tanks into the experimental tanks and acclimatized over night prior to each trial. In order to control for variation in the stress status of the fish and for stress introduced through the experimental procedure, two control groups were used. Untreated, unhandled control fish (C-1) from stock were netted from holding tanks and immediately killed by a sharp blow to the head and blood samples were drawn. For water-treated controls (C-2), two replicates were handled in an identical way to that used for the fish treated with CO, except that the feeder tanks were filled with seawater without added CO.

At the start of the treatment, the feeder and trial tanks were connected with a flexible tube (100 mm diameter), and the feeder tank was raised with a forklift to create adequate water flow. The water within the trial tank was gently reduced to approximately 200 L before addition of treated water. Upon opening of the valve in the connecting hose, the volume of water in the trial tank increased to 1,000 L in <5 min ensuring a rapid application of the treatment, in contrast to the gradual increase by Bjørlykke et al. (2011). The experiment was conducted with a randomized order of controls and treatments. After 30 min exposure to the treatment, individual salmon were rapidly netted from the tank, before being manually percussively stunned and gill cut. For operator safety, air concentrations of CO were monitored by use of portable gas detectors (GasBadge Pro, Oakdale, PA, USA) at the site of CO diffusion and during the trials. All experimental procedures were approved by the Norwegian board for research of animal use ( and were carried out according to national legislation for use of research animals.

Behavioural analysis, plasma cortisol, muscle pH and rigour index

During the experiment, the behaviour of CO exposed salmon was recorded by video, and real time descriptions were recorded continuously in writing. At the end of the CO exposure, the responses to stimulation and clinical reflexes of each individual fish were evaluated (scores of 0, 1 or 2) based on the responses to handling and eye roll (VOR) and breathing parameters of Kestin et al. (2002). A score of 0 indicates unconsciousness and is described by lack of responses to handling, eyes fixed relative to head and no or slow opercula movements, whereas a score of 2 indicates consciousness shown by immediate escaping as a response to handling, eyes roll relative to head and the fish attempt to remain upright, and regular opercula movements, with score 1 intermediate between these. However, all behavioural observations have to be judged by caution, as these can be difficult to assess.

For all fish muscle, pH and rigour index (%) were measured at the time of death, and after 6, 20, 24, 48, 60 and 66 h, according to previously published methods (Bito et al. 1983; Bjørlykke et al. 2011). Blood samples were taken from the caudal vessel at the time of death. To prepare for plasma cortisol analysis, heparinized blood (1 mL) was centrifuged (5,000 rpm for 4 min) and plasma was collected and stored on ice for 5 h and frozen at –80 °C. The measurement of plasma cortisol was carried out according to the manufacturer’s instructions using IBL’s Cortisol ELISA kit (IBL International, Hamburg, Germany). Statistical analysis of cortisol data was carried out in R (v). 2.13; (R Development Core Team 2010), using pairwise t test with p values adjusted according to Holm (1979).

Blood, brain and S. vasculosus (SV) extracts

Whole brains and SVs were used to prepare two different protein extracts. Whole salmon brains and SV were immersed in 1 × PBS (Phosphate Buffered Saline) pH 7.4 and homogenized. After high-speed centrifugation (20,000 rpm for 20 min, 4 °C), the cleared extracts (20 mg/mL) were analysed immediately or stored frozen at −80 °C until use.

Immunoprecipitation and Western blotting analysis

Immunoprecipitation was carried out using Protein A/G Plus-Agarose (Santa Cruz Biotechnology) essentially as described in Bjørndal et al. (2006). The extracts were pre-cleared for proteins with unspecific binding to agarose by incubating with 10 μL Protein A/G Plus-Agarose for 1 h at 4 °C with gently shaking. The agarose beads were removed by centrifugation, and the supernatants collected and incubated with 2.0 μg rabbit Ngb human antibody (Santa Cruz Biotechnology, sc-30144) at 4 °C with gentle shaking. After 4 h, 15 μL Protein A/G Plus-Agarose was added, and the incubation continued for additional 16 h. The agarose beads with the bound antibody complexes were collected by centrifugation and washed with PBS. Pellets were suspended in 20 μL of PBS and analysed immediately or stored at −80° C until use.

SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was carried out essentially according to Laemmli (1970). Human recombinant Ngb (50 ng/μL; PromoKine, C-60210) was used as a positive control. For Western blotting analysis, proteins separated on SDS-PAGE were transferred to PROTRANR Nitrocellulose Transfer Membrane (Whatman GmbH, Germany). Following transfer, non-specific binding sites of the membrane were blocked by incubation for 1.5 h at room temperature with 5 % non-fat dry milk and 5 % BSA in PBS. The membranes were then incubated with polyclonal rabbit anti-human Ngb (Santa Cruz Biotechnology, sc-30144) for 24 h at 4 °C using a diluted 1:50 antibody solution in PBS with 1 % (w/v) non-fat dry milk. After washing, the membranes were incubated with horseradish peroxidase linked donkey anti-rabbit IgG antibody (GE Healthcare, UK). Ngb was visualized by chemoluminescence and recorded on an Amersham Hyperfilm™ ECL.

Absorption spectrum

Absorption spectrums were measured on blood and tissue from SV. Collected blood was diluted in 50 mM potassium phosphate buffer pH 6.9. The SVs were dissected from the brains and homogenized in the phosphate buffer and centrifuged to clear the resulting solutions. Absorption spectrums were run on Agilent 8453 spectrophotometer (Agilent Technologies, Waldbronn, Germany). Reduction was obtained by grains of sodium dithionite and oxidation by grains of potassium ferric cyanide, added directly to the cuvette.


Behavioural analysis, plasma cortisol, muscle pH and rigour index

When the salmon were subjected to a sudden increase in CO levels by the influx of water with CO-H or CO-M concentrations, the fish showed a clear reaction towards the influx of water by increased swimming activity and exploration of currents source. However, there were no initial signs of fear or aversion, and identical behaviour was present when ordinary seawater entered the control tank. In both groups, all of the fishes showed escape behaviour including rapid swimming and swimming in the surface following 2–4 min of CO exposure. After approximately 5 min, all fish exposed to CO had lost equilibrium and were lying on the bottom of the tank. For CO-H, an additional period of high activity happened shortly after loss of equilibrium, lasting about 2 min (6–8 min after start of exposure). Following this period of activity, all fishes were laying on the bottom with only weak and sporadic activity such as single strokes with the tail. For the CO-H treatment, no swimming activity was recorded after 12 min of exposure. For CO-M, there were no second high activity period, but the fish had similar sporadic activities with weak movement as in the CO-H-treated fish. Also, the activity continued for a longer time in the CO-M group and was recorded for up to 19 min after start of CO exposure. In both groups, CO induced rapid strong ventilation after 2–4 min of exposure. After loss of equilibrium, the ventilation activity became slower again, until only gasping (single strong ventilation stroke with gills followed by extended periods with no ventilation activity) was seen. In the CO-M group, strong ventilation was seen after 12 min, followed by irregular gasping for the remaining time of treatment. For the CO-H group, a similar pattern was obvious, but occurred earlier with gasping evident from approximately 10–12 min after start of exposure. After 30 min of exposure, the fish were netted, and consciousness was evaluated based on the response to handling, VOR (eye roll) and breathing according to Kestin et al. (2002). For CO-H, all fishes in the first tank were given scores of 0, 0 and 1, whereas all fishes in the second tank were given scores of 0, 0, 0, for response to handling, VOR and breathing, respectively. The CO-M-treated fish showed higher variation of scores within and between tanks. Overall, the CO-M treatment gave a score of 1, 1, 1, for response to handling, VOR and breathing, respectively. However, individual fishes scored 2 or 0 on one of the parameters, in combination with a score of 1 on both of the other parameters.

Salmon exposed to both CO-M and CO-H had earlier onset of rigour mortis and also an earlier decrease in muscle pH post-mortem than controls (Fig. 1). Salmon exposed to CO had a fast onset of rigour mortis, and full rigour had developed after 6 h in the CO-H group (Fig. 1f). CO exposure resulted in a slightly lower final pH than the control group. The decline in pH and onset of rigour (Fig. 1c, d) for CO-M are slightly slower and longer, respectively.
Fig. 1

Mean pH ± SD and mean rigour index ± SD for control (water-treated) salmon (a, b), salmon exposed to CO-M (c, d) and CO-H (e, f). The two groups were percussive stunned and gill cut after exposure to CO. The fish were placed in bins of ice slurry between the measurements. n = 8 for each group

Plasma cortisol levels (Table 1) were lowest in the untreated unhandled control group (C1), and highest in the water-treated control group (C2). Both CO concentrations gave plasma cortisol levels intermediate between the two control groups, with the highest variation in the CO-M group. We regard cross-reactivity to be similar and low in both groups.
Table 1

Plasma cortisol (ng/mL) in salmon after 30 min of CO treatment



Median (ng/mL)

MAD (ng/mL)

p values


































CO was administered in high (CO-H) and medium (CO-M) concentrations. Control groups were untreated fish (C-1) and water-treated fish (C-2). Cortisol levels are given in ng/mL. n = 8 in each group. Statistical differences are indicated by p values between all groups. Estimates of centre and spread are given by median and MAD (median absolute deviation)

Neuroglobin detection in salmon brain by Western blotting analysis

A semi-quantitative Western blotting analysis was applied to analyse Ngb protein levels in the SV and whole brain. Because of the low concentrations, Ngb had to be immunoprecipitated from the brain extracts in order to give adequate signal. As shown in Fig. 2, immunoprecipitated samples from three whole brains and five SV were detected with the expected mass of 17 kDa (Burmester et al. 2000). As an estimate, by comparison with an internal standard of recombinant human Ngb, the content of Ngb in salmon brain was found to be in the range of 1–10 μg/g brain tissue.
Fig. 2

Western blot analysis of Ngb protein tissue expression. Immunoprecipitated extracts of whole brain and SV, lane 1 and 4, respectively, and extracts of whole brain and SV not immunoprecipitated, lane 2 and lane 3, respectively. Recombinant human Ngb was applied as a positive control, lane 5

Absorption spectra

A comparison between absorption spectra of blood and cleared supernatant of SV showed that they were identical, indicating that the main coloured protein in the SV samples was haemoglobin (Fig. 3). The binding of CO after reduction with dithionite showed peaks at 539 and 568 nm.
Fig. 3

Spectra of haemoglobin from Saccus vasculosus and blood. Oxidized with potassium ferric cyanide (K3Fe(CN)6), reduced with dithionite (Na2S2O4), and after addition of CO


The uptake of CO in fish has not been studied in detail. However, similar to animals where it has been studied, CO uptake will vary greatly depending on the concentration of CO but also on the activity and physical condition of the animal. There is a definitive need for control of the CO concentration during experimental trials. However, measuring CO in water is more technically demanding compared to air measurements. In the presented trials, we have estimated the concentrations of CO based on different time of diffusion into water. Preliminary results from CO in water analysis have shown that different diffusion times give different saturation levels of CO. Also, treatments of fish using different estimated levels show different behavioural effects and time to loss of equilibrium, indicating that there are true differences in levels of CO in the water between the different experimental tanks.

In the present study, salmon were rapidly exposed to CO-H or CO-M, in contrast to the gradual increase used by Bjørlykke et al. (2011). As reported in the previous study, the fish did not seem to have aversive reactions to the CO gas as such. The initial response to the treatment was calm with exploration of the current source. The increased activity of the fish in the first 2 min after start of addition of water was most likely attributed to swimming and positioning in the induced turbulence. This indicates that CO gas is not aversive to salmon, similar to other animals including humans. Hence, the observed escape behaviour and active surface seeking seen after 2–4 min are more likely to originate from secondary hypoxia sensing mechanisms since CO very effectively will replace O2 and inhibit its use throughout the fish body due to its much higher affinity for oxygen binding proteins than oxygen itself (Blumenthal 2001; Goldstein 2008). Both CO-H and CO-M are likely to cause a near complete and near non-reversible displacement of O2 in the fish. This will rapidly cause total anoxia in all tissues inducing a range of physiological changes of which many may be defined as oxygen-sensing mechanisms (Renshaw and Nikinmaa 2007). Examples of secondary effects that may signal hypoxia in this setting are acidosis due to anaerobic metabolism that increases the lactate concentration, decreased ATP or increased ROS production. All of these are putative oxygen-sensing mechanisms (Lahiri et al. 2006) and may elicit strong aversive reactions. From previous studies (EFSA 2009; Mejdell et al. 2010), it is known that CO2 induces panic by causing acidosis in the fish, and preliminary work (unpublished results) shows that addition of acid (HCl) induces similar aversive responses.

Due to the high affinity of CO haemoglobin, as well as its O2 utilization blocking effects at high concentrations, it is likely that the early onset of rigour mortis and rapid pH decrease is caused anaerobic metabolism and lactate production (Fig. 1). The CO treatment in the present study showed a clear effect on the onset of rigour mortis compared to the findings of Bjørlykke et al. (2011). This tells us that the CO treatment presented here resulted in a higher degree of oxygen exclusion than in the previous work. This indicates that there are differences in the effect of CO related to method of administration (gradual vs. abrupt increase in concentration) and/or the amount of oxygen present together with the CO gas. pH measurements showed a similar variation in the two groups (control and CO-H). However, the initial pH is lower in the CO-treated groups compared to control indicating a rapid onset of anaerobic metabolism due to O2 deficiency.

Stress load was assessed by measuring plasma cortisol. Cortisol is one of the major mediators of external stimuli and frequently used as a marker of stress (Mommsen et al. 1999). In our experiments, the water-treated control showed elevated levels of cortisol implying that aspects of the experimental procedure other than the presence of CO were perceived as stressful. This highlights the difficulty of assessing stress in fish as only minor changes in the environment may elicit quite strong cortisol responses. Interestingly, the CO-treated fish were intermediate between the untreated unhandled control and the water-treated control with respect to cortisol response. This indicates that the stress induced by the experimental procedure was to some degree alleviated by the CO treatment. In our opinion, this is an indication that CO reduces the impact of the stress and is in support of our hypothesis that CO may be used as a calming and/or sedative agent during slaughter of Atlantic salmon.

In order to efficiently kill fish by hypoxia, tissue O2 needs to be depleted rapidly, especially in the brain, to cause swift unconsciousness and brain death, so as to minimize adverse welfare effects. Both the SV and Ngb have been implicated in oxygen storage and/or damage protection in the brain during O2 deficiency (Burmester and Hankeln 2009). Hence, we hypothesized that these oxygen stores are important during hypoxia in salmon, and that brain function would cease more rapidly if these could be rapidly depleted of oxygen. Furthermore, we believe that this would have a calming and/or sedative effect on the fish and be beneficial to welfare. CO has a high affinity for haemoglobin and efficiently replaces O2 due to its >200-fold higher affinity for haeme groups (Roughton 1970). Hence, our hypothesis is that CO effectively removes oxygen from the blood (Hb), muscle tissues (Mb) and other oxygen storages including SV. Thus, for the hexa-coordinated neuroglobins, the in vivo affinities for O2 and CO are in the same range (Dewilde et al. 2001; Kiger et al. 2004) and may exert a different role in the metabolism of O2 than haemoglobin.

Western blotting analysis showed that Ngb is present both in the brain and in the SV of salmon. From the data in Fig. 2, we estimated that the amount of Ngb in the brain and SV was between 1 and 10 μg/g of salmon brain. The low amount of sNgb indicates that Ngb may serve a function as a regulatory O2 sensor linked to signalling pathways that respond to changes in O2 concentrations—for example, hypoxia. During our studies, we have observed that the size of SV varies (data not shown), and we hypothesized that this might be due to differences in the degree of hypoxia to which the fish have been exposed that may induce growth of SV as well as increasing transcription of Ngb. Ngb has been shown to be inducible under hypoxic conditions in turtle and in zebrafish (Milton et al. 2006; Roesner et al. 2006) and might be sensitive to variations in the O2/CO concentrations in the salmon.

It is important to consider the product quality of fish since CO enhances the passage of blood cells into the surrounding body in addition to the bright colour of carboxymyoglobin formed (Bjørlykke et al. 2011). CO is taken up through the gills and transported by the blood to the SV. On cutting the gills, both blood and gills had a clear red colour. The spectra in Fig. 3 show that colour of the blood and the SV was similar, indicating that the main pigment in SV is Hb. The high concentration of Hb in the SV indicates that it is an oxygen storage reservoir (Sanson 1998).


The technical staff at the Institute of Marine Research facilities at Austevoll and in Matre are thanked for their assistance. We also thank Åse Spangelo at Yara Praxair, Porsgrunn, for supply of CO gas and gas equipments. This project was sponsored by Norwegian Research Council (NFR), Project Number: 190021/S40.

Copyright information

© Springer Science+Business Media Dordrecht 2012