Interspecific variation in hypoxia tolerance, swimming performance and plasticity in cyprinids that prefer different habitats

Oxygen availability in aquatic habitats is a major environmental factor that influences the ecology, behaviour and physiology of fish (Martínez et al., 2011). Hypoxia [defined as any level of dissolved oxygen (DO) low enough to negatively impact the behaviour and/or physiology of an organism] occurs naturally in many aquatic systems (Pollock et al., 2007). However, the frequency and extent of hypoxia is increasing associated with anthropogenic activities such as eutrophication and pollution of water bodies (Rabalais et al., 2010). Therefore, it has become increasingly important to understand the mechanisms that fish use to persist and survive under hypoxic conditions and to identify indicators of hypoxia tolerance that can be compared across species and systems. In this study, we use a closely related group of cyprinid fishes from China to explore habitat-associated hypoxia tolerance, relationships among tolerance indices and mechanisms that may contribute to the interspecific patterns observed.

Critical oxygen tension for routine oxygen consumption rate (P_crit) is the minimum oxygen level required to sustain routine oxygen consumption rate (Ṁ_O2,rout), and is considered to be an indicator of an animal's hypoxia tolerance (Ultsch et al., 1978; Mandic et al., 2009). In addition to P_crit, the oxygen threshold for the loss of equilibrium (LOE), which represents the partial pressure of oxygen at which the fish can maintain balance, is also a frequently used indicator of hypoxia tolerance (Barnes et al., 2011; Mandic et al., 2013). Aquatic surface respiration (ASR), whereby fish breathe water from surface film, is a common response of water-breathing fish to extreme hypoxia (Shingles et al., 2005; Sloman et al., 2006). ASR is hypothesised to be triggered by environmental oxygen tensions at which respiratory mechanisms fail to compensate for environmental hypoxia (Takasusuki et al., 1998). Therefore, the oxygen threshold for ASR is another potentially useful hypoxia tolerance indictor in addition to P_crit and LOE.

In many fish species, swimming performance is postulated to be a central determinant of Darwinian fitness (Brett, 1964; Plaut, 2001; Blake, 2004). In fish, the determination of maximum sustainable swimming speed or critical swimming speed (U_crit) is widely used to evaluate aerobic swimming performance (Gregory and Wood, 1998; Plaut, 2001; Lee et al., 2003; MacNutt et al., 2004). Aerobic swimming performance may be limited by either oxygen uptake and delivery or the aerobic metabolic capacity of the muscle. Ecologically, decreased swimming performance, which includes both swimming speed and swimming efficiency, in hypoxic water may render animals more vulnerable to predation and may also affect the foraging efficiency of predators (Abrahams et al., 2007). Therefore, the mechanism by which fish can maintain their swimming ability under hypoxic conditions may be closely related to their survival in such an environment.

The ability of fish to maintain their swimming ability under hypoxic conditions may be affected by acclimation to hypoxia, i.e. acclimation may induce a phenotypically plastic response that improves tolerance to hypoxic stress. For example, the maximum metabolic rate (Ṁ_O2,max) and swimming performance of goldfish (Carassius auratus) improved significantly after acclimation to hypoxia for 48 h (Fu et al., 2011). And in several species of cyprinids, including the crucian carp (Carassius carassius) (Sollid et al., 2003), goldfish (Sollid et al., 2005; Mitrovic et al., 2009) and scaleless carp (Gymnocypris przewalskii) (Matey et al., 2008), hypoxia exposure has been found to result in dramatic changes in gill morphology (Sollid et al., 2003) that reduces the water–blood diffusion distance (Matey et al., 2008).

Hypoxia tolerance and physiological plasticity may differ between fish that live in rapid-flow habitats and fish that live in slow-flow habitats. Rapid-flow rivers often exhibit little DO change and generally have high DO levels, whereas some small, isolated bodies of water, such as ponds, exhibit large daily DO fluctuations. The Cyprinidae is one of the largest families of vertebrates in the world. This family has a wide geographic distribution, including mainland Eurasia, Japan, the East Indian Islands, Africa and North America. There are ~532 species of cyprinid within ~132 genera in China, and the phylogenetic relationships among these fish are well documented (Wang, 2005). Because of considerable morphological and physiological diversity, cyprinids exist in a wide variety of habitats (Howes, 1991). Therefore, we investigated the hypoxia tolerance, swimming performance and plastic response to hypoxia in 12 cyprinid species that live in slow-, intermediate- or rapid-flow habitats (Table 1): Schizothorax prenanti (Tchang 1930) (SP), Onychostoma sima (Sauvage & Dabry de Thiersant 1874) (OS), Spinibarbus sinensis (Bleeker 1871) (SS), Carassius auratus (Linnaeus 1758) (AA), Carassius carassius (Linnaeus 1758) (CA), Cyprinus carpio Linnaeus 1758 (CC), Aristichthys nobilis (Richardson 1845) (AN), Hypophthalmichthys molitrix (Valenciennes 1844) (HM), Parabramis pekinensis (Basilewsky 1855) (PP), Ctenopharyngodon idellus (Valenciennes 1844) (CI), Ctenopharyngodon piceus (Richardson 1846) (CP) and Zacco platypus (Temminck & Schlegel 1846) (ZP). The objective of this study was to test whether hypoxia tolerance, gill remodelling ability, swimming performance and plasticity are related to habitat. To test our hypotheses, we used both traditional statistical analyses and phylogenetically independent contrasts (PIC).

The effect of species on ASR₅₀ was significant, while neither habitat nor fish length (covariate) were significant (nested ANCOVA; Fig. 1A). The effect of habitat on ASR₅₀ was also not significant when controlling for phylogeny (PIC, P=0.429).

Both habitat and species showed significant effects, while body length showed no effect on LOE₅₀ (nested ANCOVA; Fig.1A). The effect of habitat on LOE₅₀ was also significant when controlling for phylogeny (PIC, P=0.004). The LOE₅₀ of fish species from rapid-flow water exhibited significantly higher LOE₅₀ values than the intermediate-flow group, while the latter exhibited significantly higher LOE₅₀ values than the species from slow-flow habitats (one-way ANCOVA, P<0.001; Fig. 1B). Neither AA nor CA showed any sign of LOE₅₀ after exposure to DO-free water for 1 h.

The ASR₅₀ value was positively related to LOE₅₀ in all 12 species (R²=0.580, P=0.004; Fig. 1C).

P_crit varied from 0.78±0.05 in AN to 3.00±0.38 mg O₂ l⁻¹ in ZP (Fig. 2A), whereas routine metabolic rate (Ṁ_O2,rout) varied from 29.5±2.1 mg O kg⁻¹ h⁻¹ in PP to 63.6±4.7 mg O₂ kg⁻¹ h⁻¹ in SP (Fig. 2B). There was no significant difference in either P_crit or Ṁ_O2,rout among habitats (nested ANCOVA, P>0.05). However, the habitat effect was significant for P_crit when controlling for phylogeny (Fig. 2A,B). One-way ANCOVA indicated a significant difference in P_crit between the rapid-flow group and the other two groups. The mean P_crit value was positively related to mean Ṁ_O2,rout in all 12 species (R²=0.341, P=0.038; Fig. 2C).

In the control group (normoxia), the fish species showed great variation in U_crit, ranging from 7.05 body lengths (BL) s⁻¹ in OS to 3.62 BL s⁻¹ in CI (Fig. 3A). Both species and body length showed significant effects (P<0.001) while habitat showed no significant effect on U_crit (nested ANCOVA; Fig. 3A). However, the effect of habitat on U_crit was shown using phylogenetic analysis (PIC, P=0.046). The U_crit of the rapid-flow fish species was significantly higher than the other fish groups (one-way ANCOVA, P<0.001).

Under hypoxic conditions, all fish species showed significantly lower U_crit values (P<0.001). The U_crit value of control fish decreased by 21–60%, with the percentage reduction being greater in fish from rapid-flow habitats (significant species by condition interaction, P<0.001; Fig. 3B).

After 48 h hypoxia acclimation, only CA and AA showed significantly higher U_crit values under normoxic conditions, whereas both SP and PP showed significantly lower U_crit values compared with the control group (P<0.05; Fig. 3). However, when measured under hypoxic conditions, half of the fish species from the intermediate-flow group and all four fish species from the slow-flow group showed significantly higher U_crit values compared with the control fish under hypoxic conditions. Therefore, the hypoxia acclimation had no effect on the U_crit measured under hypoxia for the rapid-flow species, but significantly improved the U_crit measured under hypoxia for the slow-flow species.

Ṁ_O2,max also showed great variation among the different species. SP showed the highest Ṁ_O2,max value, with 267 mg O₂ kg⁻¹ h⁻¹, whereas CA showed the lowest value, with 117 mg O₂ kg⁻¹ h⁻¹ (Fig. 4A). However, neither habitat nor body mass showed significant effects on Ṁ_O2,max (nested ANCOVA; Fig. 4A). The effect of habitat on Ṁ_O2,max was also not significant when controlling for phylogeny (PIC, P=0.142).

Under hypoxic conditions, all fish species showed significantly lower Ṁ_O2,max values (P<0.001). The Ṁ_O2,max value of control fish decreased by 53 to 81%, with the percentage reduction greater in fish from rapid-flow habitats (significant species by condition interaction, P<0.001; Fig. 4B).

After 48 h hypoxia acclimation, only CA showed significantly higher Ṁ_O2,max values under normoxic conditions, whereas AN showed significantly lower Ṁ_O2,max values compared with the control group (P<0.05; Fig. 4). However, under hypoxic conditions, half the fish species from the intermediate group and three of four fish species from the slow-flow group showed significantly higher Ṁ_O2,max values compared with the control fish (P<0.05). Hypoxia acclimation had no effect on the Ṁ_O2,max for any of the four rapid-flow fish species.

All four fish species that were evaluated exhibited no gill morphology changes after hypoxia acclimation, as indicated by the protruding lamella height. The protruding lamella heights were 62.1±4.2 μm (N=4 for all four species both before and after hypoxia acclimation) versus 54.5±2.9 μm, 47.0±3.1 versus 37.6±3.3 μm, 51.8±2.0 versus 53.8±1.3 and 75.2±2.1 versus 62.0±3.8 μm in the hypoxia-acclimated and non-acclimated SP, OS, SS and ZP, respectively.

The primary objective of this study was to answer the question of whether hypoxia tolerance and swimming performance differed in fish species that live in different habitats. Habitat-specific differences in hypoxia tolerance and swimming performance were detected by PIC, while only hypoxia tolerance (as suggested by LOE₅₀) was detected by traditional contrasts in this study, suggesting that PIC is a more powerful test of habitat than the traditional analyses. The fish species from rapid-flow habitats showed lower hypoxia tolerance but better swimming performance and greater sensitivity to changes in DO compared with fish from slow-flow habitats, as expected. Furthermore, the fish from the slow-flow habitats showed improved swimming and respiratory capacities when measured under hypoxic conditions after hypoxia acclimation, possibly related to changes in gill morphology, whereas none of rapid-flow fish showed any improvement in aerobic swimming performance after hypoxia acclimation.

Habitat-specific differences in hypoxia tolerance as suggested by LOE₅₀ and P_crit were detected by PIC in this study. However, the habitat-specific difference in P_crit was not shown by traditional analyses. The difference between these two methods is that PIC take phylogenetic relationships into account. It is suggested that PIC may be more powerful than traditional nested ANOVA. It has been demonstrated that P_crit is quite changeable and is affected by routine metabolic rate and, therefore, nutritional status (Hochachka, 1986; Guppy and Withers, 1999), hypoxia acclimation (Fu et al., 2011) and temperature (Barnes et al., 2011). A positive relationship between P_crit and Ṁ_O2,rout was also shown in the present study (Fig. 3C). Another important reason that P_crit may not be an appropriate indicator of hypoxia tolerance is that it neglects the role of anaerobic metabolic capacity in hypoxia tolerance (this is particularly true for both AA and CA in the present study).

An animal's ability to tolerate environmental fluctuations requires the integration and coordination of behavioural, physiological and biochemical processes (Sloman et al., 2008). Behavioural responses such as ASR under hypoxic conditions may be beneficial for surviving hypoxia; however, the threshold for initiating ASR may be affected by perceived predation risk (Sloman et al., 2008). Although ASR may be predicted to be triggered by environmental DO levels at which the respiratory mechanisms fail to compensate for environmental hypoxia (Takasusuki et al., 1998), other studies have also supported the hypothesis that there is an element of flexibility in the performance of ASR (e.g. Sloman et al., 2008). In the present study, there were no significant differences in the ASR among fish with different preferred habitats. Furthermore, both AA and CA, the champions of anoxia tolerance, showed a relatively higher ASR threshold, suggesting that ASR may not be an appropriate hypoxia tolerance indicator for these two fish species (Fig. 2). This may be because some fish species, such as AA and CA in this study, performed ASR early whereas other fish did not. We can also speculate that ASR may change as a result of predator stress variation.

Nevertheless, this study suggested that the LOE might be a better indicator of hypoxia tolerance, whereas P_crit and ASR may not be appropriate for some species because of physiological and behavioural factors.

Cyprinids living in rapid-flow habitats showed higher U_crit values compared with slow-flow fish species, independent of phylogenetic relationships, which is consistent with our hypothesis. Similar to P_crit, traditional contrasts showed no significant difference among fish species from different habitats, again suggesting that PIC was more powerful in terms of detecting the habitat difference than traditional analyses. Furthermore, the swimming performance of the rapid-flow fish was more sensitive to changes in DO. This may reflect more stable and high DO conditions in rapid-flow habitats. By contrast, it may be critical for slow-flow fish species to maintain the majority of swimming performance under hypoxic conditions to survive in such an environment.

We determined whether swimming performance showed adaptive plasticity after hypoxia acclimation and whether such plasticity differed among groups of fish species from different preferred habitats. All four rapid-flow fish species showed no improvement in U_crit after hypoxia acclimation either under normoxic or hypoxic conditions. However, in the slow-flow fish species, all four fish species showed improved swimming performance under hypoxic conditions, whereas AA and CA also showed improved swimming performance under normoxic conditions. Similar to swimming performance, the fish species from slow-flow habitats also showed significant gill morphology changes whereas fish species from rapid-flow habitats showed less or no gill plasticity after 48 h hypoxia acclimation. The improved swimming and respiratory capacities after hypoxia acclimation might be partially due to the gill morphological change. Interestingly, we provide evidence here that gill remodelling may not be phylogenetically dependent because closely related fish, such as and PP and CI, showed alternative morphological responses after hypoxia acclimation. We further suggest that swimming performance and gill flexibility are habitat specific as long-term adaptions to the habitat DO condition. Besides morphological changes, physiological mechanisms might also contribute to the improved swimming performance of hypoxia-acclimated fish. It has been found that hypoxia acclimation results in an increase in haemoglobin concentration and blood oxygen carrying capacity (Wells et al., 1989; Silkin and Silkina, 2005), an increase in the number of muscle mitochondria and muscle myoglobin concentration, and a higher capillarization of muscle, which improves the extraction and utilization of circulating oxygen stores at low DO levels (Sänger, 1993).

The acclimation effect on U_crit was more profound when fish were tested at low DO levels. Increased oxygen uptake capacity, which may have been elicited by hypoxia acclimation, may not result in increased swimming performance under normoxic conditions because the limiting factor is not likely to be the availability of oxygen. However, when measured under hypoxia, slow-flow fish showed significant swimming plasticity whereas rapid-flow fish did not.

In conclusion, this study clearly demonstrated that there was a phylogenetically independent habitat-specific difference in both hypoxia tolerance and swimming performance among fish from habitats with different flow conditions. This difference may reflect responses to flow regime and associated differences in DO among habitats. Rapid-flow fish showed poor hypoxia tolerance but stronger swimming performance than the slow-flow fish, as expected. The slow-flow fish also displayed significant gill morphology changes and improvement in swimming performance after hypoxia acclimation, whereas rapid-flow fish showed no such ability. Furthermore, the swimming performance of the slow-flow fish was less sensitive to changes in DO compared with the rapid-flow fish. The differences in sensitivity and plasticity of swimming performance in the fish from habitats with different flow conditions may be due to differences in oxygen fluctuation among the different habitats. Our results also suggest that PIC might be more powerful than traditional statistical approaches when detecting habitat differences in both swimming performance and hypoxia tolerance.

Juveniles of the 12 cyprinid species (Schizothorax prenanti, Onychostoma sima, Spinibarbus sinensis, Carassius auratus, Carassius carassius, Cyprinus carpio, Aristichthys nobilis, Hypophthalmichthys molitrix, Parabramis pekinensis, Ctenopharyngodon idellus, Ctenopharyngodon piceus and Zacco platypus) were either collected by local fishermen or caught by hook-and-line angling from a local river or lake, except goldfish, which were bought from a local market in Chongqing City in southwest China. Based on our data on local fishery catchment and water velocity where fish were collected, we classified the 12 species into three groups: rapid flow, intermediate flow and slow flow (Table 1). The fish were maintained in a re-circulating-water rearing system at Chongqing Normal University for at least 2 weeks prior to experimentation. During this time, the temperature of the de-chlorinated freshwater was maintained at 15.0±0.5°C and the oxygen content was maintained above 10 mg l⁻¹. The photoperiod was 12 h:12 h light:dark. One-tenth of the water was replaced daily with freshwater to maintain good water quality. Throughout the experimental period, the fish were fed daily to satiation with commercial forage until 48 h before the experimental trials. Fish were used only once in any experimental trial. All procedures were conducted in accordance with the national animal regulations.

To quantify hypoxia tolerance, we determined the oxygen tension at which individual fish reached ASR₅₀ thresholds and LOE₅₀ thresholds during the same experiment. Briefly, for a given species, three groups of 10 individual fish were transferred from the holding tank to a 30 l tank and held under continuous slow-flow-through conditions (for maintenance of water quality) for 4 h prior to the experiment. At the start of the experiment, a mesh screen was placed below the waterline to prevent the fish from accessing the water–air interface. Inflowing water was shut off and nitrogen gas was introduced into the tank to rapidly decrease DO from normoxic levels of ~10 mg l⁻¹ to 5 mg l⁻¹, 2.5 mg l⁻¹ and then 1.2 mg l⁻¹. Thereafter, the oxygen tension was decreased in the same step-wise manner but in smaller steps of 0.1 mg l⁻¹ to a final DO level of 0 mg l⁻¹. Fish were held for 1 h intervals at each DO level and the change in DO between steps required less than 1 min. At each DO level, the total number of attempts of ASR and LOE were counted over 20 successive 3 min intervals. An attempt to perform ASR of individual fish was defined as the point where the fish made contact with the mesh surface suspended below the water–air interface and the ASR₅₀ value of any fish group was defined as the point at which five of 10 individual fish made contact with the mesh surface suspended below the water–air interface for three consecutive observations. The elapsed time between ASR of the first and the fifth fish was 40 to 80 min in all species except goldfish and crucian carp that showed relatively large variation (1 to 4 h). LOE of individual fish was defined as the failure of the fish to maintain dorsoventral orientation and the LOE₅₀ value of any fish group was defined as the DO point at which five of 10 individual fish failed to maintain a dorsoventral orientation for three consecutive observations. The elapsed time between LOE of the first and the fifth fish was 20 to 60 min in all experiment fish species except goldfish and crucian carp, which showed no LOE.

After 48 h fasting, 15 fish were randomly selected from each experimental group and placed in 160 ml respirometers for measurement of Ṁ_O2,rout and P_crit (Zhang et al., 2010). The fish were allowed to recover from transfer to the respirometer for 4 h. During this time, continuously aerated water flowed (3 cm s⁻¹) through the respirometer. Subsequently, the respirometer was closed and Ṁ_O2,rout was measured over a range of water DO values as the fish depleted the oxygen within the closed respirometer, beginning at ~95% saturation and decreasing to 1% saturation with a duration of 90 to 120 min. If the fish showed movements such as struggling and moving back and forth during the experiment, the data from the trial were discarded. To measure DO, the circulating water from the respirometer was drawn from the respirometer by a peristaltic pump, forced past a DO probe (HQ20, Hach Company, Loveland, CO, USA) housed in a sealed thermostated chamber, and then returned to the respirometer. The temperature of the system was maintained at 15±0.2°C.

The P_crit is the point at which Ṁ_O2,rout could no longer be maintained with a further reduction in the water O₂ tension and was estimated for the individual fish with the two-segment linear model described by Yeager and Ultsch (Yeager and Ultsch, 1989).

After the 2 week habituation period, the fish were fasted for 48 h, and 60 fish of similar size were randomly selected and divided into the hypoxia acclimation group and control acclimation group. The water temperature was maintained at 15°C. Thirty fish were transferred to a 120 l exposure chamber in which hypoxia was achieved by covering the surface of the water with translucent plastic and bubbling nitrogen into the water (Matey et al., 2008). The DO was reduced from aerated values of 10 to 0.3 mg l⁻¹ over 1 h and was then maintained at 0.3 mg l⁻¹ for 48 h. During this time, the water DO was continuously monitored using a DO probe.

After 48 h acclimation to hypoxic or normoxic (control group) conditions, 20 fish were selected from each group and subjected to a critical swimming speed (U_crit) test in either normoxic water (10 mg l⁻¹, N=10) or hypoxic water (1 mg l⁻¹, N=10). The water DO ranged from 10.2 to 10.4 mg l⁻¹ for normoxic swimming conditions and from 1 to 1.2 mg l⁻¹ for hypoxic swimming conditions.

The gill morphology of only four fish species was evaluated, S. prenanti, O. sima, S. sinensis and Z. platypus, because the gill morphology change that was elicited by hypoxia acclimation of the other eight species has previously been performed by our laboratory (Dhillon et al., 2013) (Table 1).

After 48 h acclimation to hypoxic or control conditions, four fish from each group were immediately euthanised using neutralised tricaine methanesulphonate (MS-222, 50 mg l⁻¹) and terminally sampled. The second gill arch from the right side of each fish was removed, rinsed and immediately fixed in cold Karnovsky's fixative for scanning electron microscopy (SEM) (at the Third Military Medical University, Chongqing, China).

The middle part of each fixed gill arch (5 mm) with up to 20 filaments in both the anterior and posterior rows was used for SEM. All fixed gill tissues were rinsed in phosphate-buffered saline and post-fixed in 1% osmium tetroxide for 1 h. The gill tissues were dehydrated in ascending concentrations of ethanol from 30% to 100%, critical-point dried with liquid CO₂, mounted on stubs, sputter-coated with gold-palladium, and examined with a Hitachi S 3400 scanning electron microscope (Hitachi High-Technologies Europe, Krefeld, Germany) at an accelerating voltage of 15 kV. The protruding lamella height was measured to estimate changes in the gill morphology.

SPSS Statistics 17 (IBM, Armonk, NY, USA) was used for data analysis. P-values <0.05 were considered statistically significant, and all data are presented as means ± s.e.m.

The effects of habitat and species (nested within habitat) on hypoxia tolerance and swimming performance (we only tested the swimming performance of control group measured under normoxic conditions) were determined by a nested analysis of covariance [ANCOVA; with body size (length or mass) used as the covariate]. One-way ANCOVA followed by a post hoc Duncan's multiple-comparisons test was used to detect differences among habitats. The effect of species and DO (oxygen concentration in which the trait was measured), and their interaction on U_crit and Ṁ_O2,max was tested by two-way ANCOVA. The effect of hypoxia acclimation on U_crit and Ṁ_O2,max was detected using t-tests. For the 12 fish species, Pearson correlation was used to examine the relationship between ASR₅₀ and LOE₅₀, and P_crit and Ṁ_O2,rout.

We also conducted phylogenetically independent ANOVAs, which tested for differences in hypoxia tolerance and swimming performance (for the control group measured under normoxia) among species inhabiting environments under different flow conditions. We used the PDSIMUL and PDANOVA programs (Garland et al., 1993) to perform PICs. Using these programs, we simulated trait evolution as Brownian motion with the means and variances of the simulations set to the means and variances of the original data. We performed 1000 simulations, producing a null distribution of F-statistics against which the F-value of one-way ANOVA from the actual data could be compared to assess the statistical significance (i.e. we determined how different the observed patterns were from those expected via genetic drift alone). We constructed a best-estimate phylogenetic hypothesis for this group of species based on previous morphological and molecular studies (Fig. 5). All branch lengths were set as equal to one.