All procedures were reviewed and approved by the Animal Care Committee at the University of Manitoba, Canada (Approval ID: AUP-F11-004) under the guidelines of the Canadian Council of Animal Care. No animals were sacrificed, all efforts were taken to ameliorate animal suffering and undue stress, and there was no mortality during any of the tests.

A total of 70 juvenile A. fulvescens (body mass: 30.51±1.21 g (mean ± S.E.); age: 1+; sex: unknown) obtained from Grand Rapids Fish Hatchery (Grand Rapids, MB, Canada) were kept at 17±1°C in flow-through holding tanks at the University of Manitoba, Canada. The light regime was 12 h light: 12 h dark (12L∶12D). A. fulvescens were fed daily using a mixture of bloodworm (San Francisco Bay Brand, Newark, CA, USA) and sinking trout pellet (Martin Mills Ltd., Elmira, ON, Canada).

Four static respirometers (each 0.83 l) and a mixing pump were submerged in a 100 l opaque tank, filled with freshwater maintained at 17±0.1°C. Oxygen content (% air saturation; O_2sat) of the water in the tank was controlled using two air stones combined with a stream of nitrogen bubbles [39]. Depending on the experiment, water in the tank was maintained at an oxygenation level between 100% and 30% O_2sat.

Respirometers were made of transparent glass tubing and were designed to allow a degree of spontaneous activity of A. fulvescens, including body undulations with tail excursions>90° relative to the body axis. Respirometers were situated in a sound isolated room with no other ongoing experiments to minimize any disturbance of the fish.

Measurements of MO₂ (mg O₂ h⁻¹) were carried out every 9 min using computerized intermittent flow respirometry allowing long term (>48 h) repeated measurements [40]. Each respirometer was fitted with two outlet and two inlet ports as described previously [41]. The repeated respirometric loops consisted of a 4 min flushing phase during which a pump flushed the respirometer with ambient water through one set of ports. The second set of ports and a pump secured re-circulation of water in the respirometer in a closed circuit phase for 5 min, divided into a waiting phase (2 min) and a measurement phase (3 min).

Oxygen partial pressure was measured at 1 Hz by a fiber optic sensor (Fibox 3 connected to a dipping probe; PreSens, Regensburg, Germany) located in the re-circulated loop. The flush pump was controlled by AutoResp software (version 2.1.3; Loligo Systems, Tjele, Denmark) that also calculated the MO₂ in the measurement phase using the oxygen partial pressure and standard equations [42], [43]. Preliminary testing demonstrated that the duration of the measurement phase (3 min) ensured that the coefficient of determination (r²) associated with each MO₂ measurement was always>0.95, similar to previous studies [44]. Corrections of background respiration (i.e. microbial respiration) followed Jones et al. [45].

A. fulvescens were selected randomly and fasted for 48 h to ensure a post absorptive state prior to experimentation. Subsequently, A. fulvescens were introduced to the respirometers and acclimated for 20 h. The light regime during the fasting and acclimation periods was 12L∶12D, which included 0.5 h of gradually shifting light intensity from light to darkness and vice versa. Light intensities were 3.0 and 0.0 μmol s⁻¹ m⁻² in daylight and darkness, respectively. Starting at 16:00 h on the next day, MO₂ data were collected for the following 24 h.

Measurements of MO₂ over 24 h comprised three test groups: 1) control (100% O_2sat; 12L∶12D); 2) treatment A (30% O_2sat; 12L∶12D); and 3) treatment B (100% O_2sat; 24L). The oxygen content in treatment A (30% O_2sat) corresponded to approximately 6.2 kPa. Data collection for the three test groups was carried out in a random fashion and each test group included 10–12 individuals. After each 24 h trial, MMR_F was measured as described below.

For each test group, SMR in individual fish was estimated as the average of the lowest 10 MO₂ values collected over 24 h. This method to estimate SMR was employed because it provides measurements that are repeatable in individual fish [3].

MMR_F was measured immediately after each 24 h trial at the corresponding O_2sat level (i.e. 100% or 30% O_2sat) inside the respirometer. MMR_F was elicited using a standard chase protocol [36]. Briefly, individual A. fulvescens were transferred to a circular trough and chased to exhaustion, similar to previous studies on Atlantic sturgeon (Acipenser oxyrhynchus) and shortnose sturgeon (Acipenser brevirostrum) [46]. Upon exhaustion, identified by no further response after 5 min of manual stimulation, A. fulvescens were transferred (<20 s) to the respirometer where MO₂ measurements started immediately. MMR_F was the highest of three consecutive MO₂ measurements.

In addition, following the same chase protocol, MMR_F was measured in 36 A. fulvescens exposed to 100%, 90%, 80% or 70% O_2sat inside the respirometer. A total of 8–12 A. fulvescens were tested at each of the four O_2sat levels. Measurements of MMR_F in 100% O_2sat were repeated after 4.5 h to examine the short term repeatability of MMR_F in individual fish. These two measurements were termed initial and final MMR_F.

Finally, for each test group (i.e. control and treatments A and B), MMR_S was estimated as the single highest measurement of MO₂ (i.e. one respirometric loop) in volitionally performing individual fish during the complete 24 h trial (i.e. after acclimation). These data were used to test for correlations between MMR_F and MMR_S in individual fish (see Data analysis).

A. fulvescens in the respirometers were recorded (25 frames s⁻¹) dorsally using a UEye camera (model UI-1640SE-C-GL; IDS, Woburn, MA, USA) equipped with a CCTV lens (model HF6M-2; Spacecom, Whittier, CA, USA). The software UEye Cockpit (version 3.90; IDS, Woburn, MA, USA) was used to download recordings to a PC. Two Scene illuminators (model S8030-30-C-IR; Guangdong, China) provided infra-red light for nocturnal recordings. All recordings were synchronized with the respirometric loops (to the nearest 1 s). For each A. fulvescens, behavioural data were collected over a 45 s time interval during the measurement phase of the respirometric loop (i.e. once every 9 min.). Behavioural data included total activity (i.e. % of time moving), and the number of body undulations with tail excursions<90° or >90° relative to the body axis (i.e. body undulations min⁻¹). For each test group, behavioural data were collected over a 1 h time interval (i.e. 6–7 respirometric loops) at 16, 20, 21, 22 and 23 h. These hourly measurements were selected to record simultaneous metabolic and behavioural changes during the light-dark transition at 21 h.

MO₂ data were body mass adjusted following previous studies [47]. Metabolic rates from the three test groups were calculated over 1 h intervals [48], with two exceptions, because the light intensity was gradually changing over 0.5 h periods at 21 h and 9 h. Therefore, the two 1 h intervals associated with 21 h and 9 h were each divided into two: 0.5 h with changing light intensities and 0.5 h with constant light intensity. The compiled data were used to compare metabolic rates over 24 h within the three test groups (i.e. control and treatments A and B). Behavioural data were compiled in the same fashion.

Metabolic and behavioural variables were compared within each test group across the time interval from 16:00 to 23:00 h using a repeated measure (RM) one way ANOVA. Relationships between behaviour and metabolic rates were investigated using least squares linear regression.

To test for metabolic differences, SMR, MMR_F and MMR_S measurements were compared between the three test groups using a one way ANOVA. MMR_F data from the four oxygen treatments (100 – 70% O_2sat) were analyzed using least square linear regression to examine the effect of decreasing oxygen levels on MMR_F.

The method recommended by Norin and Malte [3] was used to examine repeatability of the MMR_F measurements. All values of MMR_F and body mass were log₁₀-transformed prior to the analysis. Mass-independent data of MMR_F were expressed as residual values using the relationship between body mass and MMR_F. Fish with higher than expected MMR_F have positive residuals and fish with lower than expected MMR_F have negative residuals. Repeatability of the two sets of residuals (initial and final) was estimated using Spearman's rank correlation coefficient (ρ) [3].

Using metabolic rate data from the three test groups, MMR_S of each individual fish was extracted to test for correlations between individual MMR_F and MMR_S. The comparison of individual MMR_F and MMR_S was carried out in the same fashion as the repeatability analysis described above.

Log₁₀ transformations of data prior to statistical analysis were employed to meet assumptions of normal distribution of data and homogeneity of variance. If the assumptions were met, ANOVA or RM ANOVA were employed depending on design as described above. If significant, the tests were followed by pairwise multiple comparisons using the Holm-Sidak method.

If data transformations did not permit the use of parametric testing, ANOVA on ranks or RM ANOVA on ranks (Friedman) were employed depending on design as described above. The tests were followed by pairwise multiple comparisons using Dunn's method to take unequal sample sizes into account.

Tests were carried out using SigmaStat 3.01 (Systat Software, San Jose, CA, USA) and SPSS 20.0 (IBM, Armonk, NY, USA). Results were considered significant if α<0.05. All values are reported as means ± S.E. unless noted otherwise.

There were no differences between test groups (i.e. control, treatments A and B) in terms of body mass and SMR measured as mg O₂ h⁻¹ (both P>0.05). Consequently, SMR data were pooled, and the relationship between log₁₀ SMR and log₁₀ body mass was described using a linear equation [3], [47]. The slope of the relationship was 1.00±0.12 indicating that a 1.0 body mass scaling coefficient was appropriate for the SMR data. A 1.0 body mass scaling coefficient is consistent with two previous studies on green sturgeon Acipenser medirostris [47], [49]. Because the 1.0 body mass scaling coefficient was appropriate for the SMR data, the same coefficient was used for the MO₂ data collected over time (Fig. 1).

MMR_F measured as mg O₂ h⁻¹ did not differ between the control group and treatment B (P>0.05), but MMR_F from treatment A was lower than both the control group and treatment B (P<0.001). To examine the relationship between body mass and MMR_F (mg O₂ h⁻¹), data collected in normoxia were combined and the relationship between log₁₀ MMR_F and log₁₀ body mass was described using a linear equation [3], [47]. The slope of the relationship was 0.90±0.05 indicating that a 0.9 body mass scaling coefficient was appropriate for the MMR_F data. Consequently, all MMR_F data were standardized to the mean body mass of 30.5 g using 0.9 as the body mass scaling coefficient. In the following, MMR_F standardized to 30.5 g is denoted MMR_F30.5.

Metabolic rates varied substantially over the 24 h periods (Fig. 1). In the control group, metabolic rate increased significantly (P<0.001) from 112 mg O₂ kg⁻¹ h⁻¹ at 20:00 h to reach a maximum of 237 mg O₂ kg⁻¹ h⁻¹ when the light went off (Fig. 1, control), indicating a dusk metabolic peak. Thereafter, metabolic rate decreased and reached 157 mg O₂ kg⁻¹ h⁻¹ shortly before daylight. The metabolic rate decreased further in daylight and reached 112 mg O₂ kg⁻¹ h⁻¹ after 3 h.

In treatment A, metabolic rate increased significantly (P<0.05) from 116 to 150 mg O₂ kg⁻¹ h⁻¹ when the light went off (Fig. 1, treatment A). Although truncated, this metabolic peak corresponded to the dusk metabolic peak observed in the control test group. Thereafter, metabolic rate decreased to 128 mg O₂ kg⁻¹ h⁻¹ at 03:00 h, and then increased to reach 139 mg O₂ kg⁻¹ h⁻¹ during the period with increasing light intensity (09:00 h). Thus, treatment A indicated two metabolic peaks; one associated with dusk and one associated with dawn. After the light went on, the metabolic rate changed little for 1 h and then decreased to 118 mg O₂ kg⁻¹ h⁻¹ (Fig. 1, treatment A).

In treatment B, the metabolic rate remained below 119 mg O₂ kg⁻¹ h⁻¹ until 02:00 h (Fig. 1, treatment B). Data showed that the dusk metabolic peak, observed in the control group and in treatment A, was eliminated by the constant light (P = 0.64). In contrast, in treatment B, the metabolic rate tended to increase at 02:00 h and continued doing so until it reached 178 mg O₂ kg⁻¹ h⁻¹ at 08:00 h (Fig. 1, treatment B). These data indicated the presence of a darkness independent increase in the metabolic rate. The increasing metabolic rate peaked around dawn, just before the light would normally come on.

Collectively, data indicated the presence of two metabolic peaks occurring over 24 h. The first metabolic peak occurred around dusk and was noticeable in the control group and treatment A (Fig. 1). The second metabolic peak occurred around dawn and was noticeable in treatments A and B (Fig. 1).

Behavioural recordings from the control group indicated that the total activity increased in darkness (Fig. 2, control), but no statistically significant differences were identified over time (P>0.05). Similarly, the frequency of body undulations with tail excursions<90° did not change significantly over time (P>0.05). In contrast, body undulations with tail excursions>90° increased significantly over time (P<0.001) (Fig. 2, control).

Data from treatment A revealed no significant changes in the total activity over time or in the frequency of body undulations with tail excursions<90° (both P>0.05) (Fig. 2, treatment A). In contrast, the frequency of body undulations with tail excursions>90° increased significantly over time (P<0.001).

Data from treatment B revealed no significant changes over time in the total activity or in the frequencies of body undulations with tail excursions<90° or >90° (all P>0.05; Fig. 2, treatment B).

The behavioural data suggested that the frequency of body undulations with tail excursions>90° (Fig. 2) could be a major driver of the increase in metabolic rate associated with dusk (Fig. 1). Regression analysis revealed highly significant (P<0.001 in all cases) linear relationships between the frequency of body undulations with tail excursions>90° and metabolic rate (Fig. 3). The coefficients of determination (r²) for the relationships varied between test groups and were 0.68, 0.64 and 0.15 for control and treatments A and B, respectively (Fig. 3). These data suggest that metabolic variation was coupled with behavioural variation.

SMR was unaffected by hypoxia (treatment A) and constant light (treatment B) (P>0.05; Table 1), and the pooled average was 92.39±2.00 mg O₂ kg⁻¹ h⁻¹. Corresponding analyses of MMR_F30.5 revealed no differences between the control and treatment B (P>0.05), whereas MMR_F30.5 from treatment A was lower than both the control and treatment B (P<0.001; Table 1). These findings showed that 30% O_2sat reduced MMR_F30.5.

MMR_F30.5 was quantified in four separate groups of A. fulvescens exposed to 100%, 90%, 80% or 70% O_2sat to estimate the effect of hypoxia on MMR_F30.5. Body mass did not differ between the four treatments (P = 0.95). MMR_F30.5 was affected by hypoxia and decreased across the range from 100% O_2sat to 70% O_2sat (Fig. 4) as revealed by the linear regression analysis (P<0.03; r²>0.94). These findings indicated that the maximum metabolic rate of A. fulvescens is sensitive to increasing levels of hypoxia.

Analysis of repeatability followed a previous study [3] and showed that measurements of residual body mass corrected MMR_F are repeatable in individual fish. Spearman's rank correlation coefficient (ρ) for the relationship between the initial and final residual MMR_F was 0.76, and the relationship was highly significant (P<0.006) (Fig. 5).

MMR_S was extracted from each 24 h trial for comparisons between test groups. MMR_S differed significantly between all three test groups (P<0.05) (Table 1). These findings showed that MMR_S was suppressed in treatments A and B, with the most pronounced effect in treatment A (Table 1). Standardizing MMR_S to a 30.5 g fish using a 0.9 body mass scaling coefficient (i.e. equivalent to MMR_F30.5) changed MMR_S values by < 1% and had no impact on the conclusions.

MMR_F and MMR_S were compared to test the hypothesis that they would correlate positively. Data showed that residual MMR_F and residual MMR_S were not correlated in the control group (P = 0.40; ρ = 0.27) (Fig. 6, control). In contrast, residual MMR_F and residual MMR_S were positively correlated in both treatments A (P<0.05; ρ = 0.69) and B (P<0.05; ρ = 0.66) (Fig. 6, treatments A and B). These data indicated that an individual with an unexpectedly high MMR_F also has an unexpectedly high MMR_S, at least when the individual is exposed to an environmental stressor, such as hypoxia (treatment A) or constant light (treatment B).