Hypoxia Performance Curve: Assess a Whole-Organism Metabolic Shift from a Maximum Aerobic Capacity towards a Glycolytic Capacity in Fish

The utility of measuring whole-animal performance to frame the metabolic response to environmental hypoxia is well established. Progressively reducing ambient oxygen (O2) will initially limit maximum metabolic rate as a result of a hypoxemic state and ultimately lead to a time-limited, tolerance state supported by substrate-level phosphorylation when the O2 supply can no longer meet basic needs (standard metabolic rate, SMR). The metabolic consequences of declining ambient O2 were conceptually framed for fishes initially by Fry’s hypoxic performance curve, which characterizes the hypoxemic state and its consequences to absolute aerobic scope (AAS), and Hochachka’s concept of scope for hypoxic survival, which characterizes time-limited life when SMR cannot be supported by O2 supply. Yet, despite these two conceptual frameworks, the toolbox to assess whole-animal metabolic performance remains rather limited. Here, we briefly review the ongoing debate concerning the need to standardize the most commonly used assessments of respiratory performance in hypoxic fishes, namely critical O2 (the ambient O2 level below which maintenance metabolism cannot be sustained) and the incipient lethal O2 (the ambient O2 level at which a fish loses the ability to maintain upright equilibrium), and then we advance the idea that the most useful addition to the toolbox will be the limiting-O2 concentration (LOC) performance curve. Using Fry & Hart’s (1948) hypoxia performance curve concept, an LOC curve was subsequently developed as an eco-physiological framework by Neil et al. and derived for a group of fish during a progressive hypoxia trial by Claireaux and Lagardère (1999). In the present review, we show how only minor modifications to available respirometry tools and techniques are needed to generate an LOC curve for individual fish. This individual approach to the LOC curve determination then increases its statistical robustness and importantly opens up the possibility of examining individual variability. Moreover, if peak aerobic performance at a given ambient O2 level of each individual is expressed as a percentage of its AAS, the water dissolved O2 that supports 50% of the individual’s AAS (DOAAS-50) can be interpolated much like the P50 for an O2 hemoglobin dissociation curve (when hemoglobin is 50% saturated with O2). Thus, critical O2, incipient lethal O2, DOAAS-50 and P50 and can be directly compared within and across species. While an LOC curve for individual fish represents a start to an ongoing need to seamlessly integrate aerobic to anaerobic capacity assessments in a single, multiplexed respirometry trial, we close with a comparative exploration of some of the known whole-organism anaerobic and aerobic capacity traits to examine for correlations among them and guide the next steps.


Introduction
Animal life is sustained by a tightly regulated oxidization of metabolic substrates to support ATP production. An animal's oxygen (O 2 ) transport cascade system conceptualizes the movement of O 2 down its tension gradient from the ambient environment (160 mm Hg at sea level) to the mitochondria (<1 mm Hg), where O 2 is consumed to generate ATP. Hence, if the ambient environment becomes hypoxic, the gradient of O 2 tension (PO 2 ) decreases which, at some point, limits the maximum O 2 flux through the O 2 transport cascade. Thus, by constraining maximum aerobic capacity, hypoxia can potentially impact an animal's lifetime fitness, as has been seen when supra-optimal temperature constrains the maximum aerobic capacity of adult migrating sockeye salmon [1] and limits their ability to reach their natal spawning areas [2,3].
Environmental hypoxia is a pressing issue. For example, our oceans have lost 150 trillion kilograms of O 2 in the last 50 years [4], with the largest loss (40%) being manifest as large O 2 -minimum zones (<80 µmol kg −1 O 2 within the water column) in the North and Equatorial Pacific Oceans [5]. To answer whether or not our current hypoxic trajectory will mean a pessimistic outlook for fishes, we will need a much better understanding of the standing individual variation of hypoxic performance across fish species. Furthermore, this level of understanding will require tools that can assess the respiratory phenotype of fishes well beyond those we routinely use. For instance, while the Permian era, oceanic deoxygenation wiped out an estimated 96% of marine fishes [6], we now know that after the subsequent reoxygenation of the oceans teleost fishes successfully radiated into almost every aquatic habitat on the planet aided by the evolution of new and very proficiency modes of tissue oxygenation [7].
Fish employ a repertoire of compensatory responses during hypoxia to defend their O 2 supply to essential tissues [8][9][10][11][12]. For example, moderate hypoxia in the short term can trigger hyperventilation [13,14], increase blood perfusion of gill lamellae [15], induce bradycardia which increases the time for O 2 diffusion at gill lamellae [16][17][18][19], increase venous tone which improves the return of blood to the heart [20,21], induce the splenic release of red blood cell stores which increases the blood O 2 carrying capacity [22,23] and prioritize blood flow to essential tissues [24,25]. Many of these responses are aimed at preventing a hypoxemic state, a reduced arterial blood O 2 content (C a O 2 ). If this is not possible and C a O 2 falls, both venous O 2 content (C v O 2 ) and maximum O 2 uptake (ṀO 2max ) will eventually also decrease because fish rarely tap their scope to increase cardiac output in hypoxia [12].
Chronic exposure to hypoxia can establish new respiratory phenotypes in some fish species, ones that improve tolerance of hypoxia. For example, they can resort to a hypometabolic state by temporarily turning off some basic activities and lowering basal tissue O 2 demand, a response manifested by some species that can tolerate the most severe form of hypoxia-anoxia [26]. Also, they can remodel certain tissues, including the gills [27][28][29], the heart [30,31] and the swimming muscles [32] to better tune the O 2 cascade to a reduced O 2 gradient. Obviously, knowledge of these adaptive strategies and mechanisms will be crucial to know which fish species might have an evolutionary edge in a more hypoxic future, but how best to holistically measure hypoxic performance traits in fishes is less clear. We believe that the situation can be improved by adopting concepts, techniques and analytical tools that have improved measurements of thermal performance curves in fish, an area that has received considerable attention recently (see reviews: [33,34]).
Currently, the two most widely used whole-organism hypoxic performance traits for eco-physiological studies are the critical dissolved O 2 (O 2crit ) and the incipient lethal O 2 saturation (ILOS). Both measurements employ well-established techniques and can be performed under field conditions. O 2crit is the ambient dissolved O 2 (DO) level below which maintenance metabolism (i.e., standard metabolic rate, SMR) cannot be sustained by oxidative phosphorylation and is supplemented by substrate-level phosphorylation. ILOS is the DO level at which a fish loses the ability to maintain upright equilibrium and represents an ambient DO that can be tolerated only for a limited amount of time, an ILOS that is probably close to a point of no return. O 2crit and ILOS mark the DO range for Hochachka's concept of scope for survival, a term coined to characterize time-limited life when SMR cannot be supported by O 2 supply.
While the comparative values of both O 2crit and ILOS are well established because they clearly vary across species, among strains and populations within a species and even among individuals, they can also show compensatory changes after hypoxic acclimation, but their ecological relevance has been questioned. Therefore, we need tools that assess hypoxic performance beyond just O 2crit and ILOS. Sub-lethal performance curves, ones that can identify biological safety margins for an animal's activities, maybe the next step forward. For example, sub-lethal toxicity tests have replaced lethality tests and sub-lethal performance measures are similarly being routinely used to assess the thermal performance of fishes. The key question we address in this review is which whole-organism performance traits might better quantify the sub-lethal effects of hypoxia? We believe an individualized hypoxia performance curve in the form of a limiting-O 2 concentration (LOC) performance curve for maximum metabolic rate (MMR) is a crucial first step. An individual-based and standardized experimental approach would be fundamentally different from a break-point analysis of the maximum capability to uptake O 2 under normoxic and hypoxic conditions using studies that might have employed different measurement protocols [35]. But before we introduce the individual LOC curve, we briefly review the ongoing debate concerning the importance of standardizing methodologies when assessing a fish's hypoxic performance.

Current Hypoxia Tolerance Assessments Explore Only a Small Portion of Hypoxic Performance Curve
A widely-used landmark of hypoxic performance is the minimum amount of ambient O 2 needed by a fish to aerobically sustain its whole-organism maintenance metabolic demands, i.e., O 2crit (% sat.), the term we use generically for this article. This ambient O 2 threshold can also be measured as the critical O 2 tension (P crit ; mm Hg) and critical O 2 concentration (C crit ; mg L −1 ), each of which has certain benefits (for example, water O 2 saturation is dependent on ambient atmospheric pressure and temperature; P crit is dependent on ambient atmospheric pressure but independent of temperature except for the effect on water vapour pressure; C crit is dependent on ambient atmospheric pressure and temperature).
Original studies used SMR to define O 2crit , in part because, by definition, a fish has no aerobic scope at O 2crit [10,11,36]. However, debate continues concerning the use of routine metabolic rate (RMR) as the baseline for determining O 2crit (e.g., [37,38]). Some have argued that an RMR-derived O 2crit has a greater ecological relevance because it incorporates the metabolic cost of 'idling' in a wild fish that would spend little, if any, time at SMR in nature [39]. Mueller and Seymour (2011) also advanced the Regulatory Index, an area under routine or active O 2 uptake (ṀO 2 ) versus DO curve above a linear trend of perfect conformity, with a larger area indicating a greater degree of regulation [40]. RMR, however, is a highly variable baseline, one that would be difficult to replicate from lab-to-lab and fish-to-fish. Levels of spontaneous activity are difficult to quantify without directly monitoring or controlling activity, although some earlier studies used drugs to immobilize fish (see [41]). In contrast, SMR is a less ambiguous baseline that can be reliably estimated using established testing apparatus, protocols and analytical tools providedṀO 2 in undisturbed fish is continuously measured over at least 48 h (e.g., [41][42][43]). Thus, a recent consensus among fish biologists is that a SMR-derived O 2crit represents a robust analytical approach with a stable baseline that increases the repeatability of O 2crit measurements and the utility of O 2crit for comparisons [44][45][46][47]. Whether or not SMR has ecological relevance will be an ongoing debate.
All the same, activity can influence O 2crit even when SMR is used as a baseline. For example, a higher activity during hypoxia correlated with a lower O 2crit in European sea bass (Dicentrarchus labrax) until the activity measured by the amount of O 2 consumption above SMR was higher than 100 mg O 2 kg −1 during hypoxia exposure (Figure 1a). The more active fish in hypoxia are, however, the calmer fish in normoxia and have a lower SMR (Figure 1b-d). Thus, the debate will be ongoing as to whether or not such relationships relate to a pace-of-life syndrome, wherein individuals of higher metabolism generally exhibit more active behaviours and physiological traits than their conspecifics with a lower metabolism [48,49].
Beyond a reliable measurement of SMR in normoxia to set the baseline for determining O 2crit , a fish'sṀO 2 must be measured during a progressive exposure to hypoxia [10,46]. RoutineṀO 2 may remain unchanged over much of the hypoxic range or increase slightly (either because the fish becomes agitated or due to an added O 2 cost of increasing external and internal O 2 convection), butṀO 2 will eventually decrease and conform to an everdecreasing amount of available O 2 ( Figure 2). The O 2crit is interpolated from a slope determined by a slope ofṀO 2 points near to when the threshold SMR (or RMR) can no longer be maintained [46]. The O 2crit slope was reiterated as an alpha slope, average of maximum slope from threeṀO 2 points and is forced through the origin of the axis, to quantify O 2 supply capacity [50]. Whether or not the alpha slope is constant across a whole range of DO remains to be tested by some rigorous studies.  [11,25,36], and the experimental LOC lines generated for groups of fish [52,53]. An LOC performance curve, therefore, bounds and sets the capacity for the potential use of O 2 for all other activities, as similarly suggested beyond the pejus temperature for warming [54], potentially including a buffer in O 2 availability before an activity is constrained [1]. Exactly how O 2 is apportioned amongst the various activities and what governs the apportioning are unknown, and so the shape and positioning of the broken lines are unknown. We define O 2crit as a minimum amount of ambient O 2 concentration at which a fish can aerobically sustain its whole-organism minimum maintenance metabolic demands, i.e., its standard metabolic rate (SMR) [10,46]. Thus, below O 2crit ,ṀO 2 conforms in some manner with O 2 availability and ATP needs for survival until incipient lethal O 2 saturation (ILOS; where death is imminent) must be met through a combination of oxidative phosphorylation, substrate-level phosphorylation, and possible suppression of some aspects of maintenance metabolism. Hochachka [55] coined the term scope for survival as the difference between SMR and the lowest sustainable rate to which metabolism may be suppressed before reaching ILOS. As an index of the scope for survival, we calculate accumulated O 2 deficit (AOD) as the integral of the area bounded by SMR and theṀO 2 measured between O 2crit and ILOS.
The O 2crit protocol (or hypoxic challenge test) can be performed using either closedflow or intermittent-flow respirometry. With closed-flow respirometry, the fish regulates the rate of induction of hypoxia and, although metabolic wastes progressively accumulate inside the respirometry chamber [56,57], this accumulation appears to have a negligible effect on the O 2crit estimate [58]. An intermittent-flow respirometry system offers a tighter regulation of the rate of induction of hypoxia and wastes are intermittently removed, a beneficial feature for determining an LOC performance curve.
Scope for survival [55] applies to life when ambient water O 2 is below O 2crit and a fish supplements maintenance metabolism with substrate-level phosphorylation [59] and/or enters a hypometabolic state (some basic metabolic needs are temporarily switched off or turned down; see review by [8]. Metabolic suppression (often coupled with cold exposure) can involve arresting ion channels, spike potentials and protein translation, which are associated with inactivity, reduced cardiac activity and suspended growth [60][61][62][63]. A lethargic state is a constraint on RMR and AMR, something can be induced by cooling and darkness [64] and is fundamentally different from a metabolic suppression.
ILOS is a widely-used landmark on the hypoxic performance curve of fishes [10,65] and is the point where ambient DO is so low that a fish loses neurological control and can no longer maintain an upright equilibrium (except perhaps for fishes that have morphological structure to hold themselves upright even when they lose neurological control). ILOS, however, is quite sensitive to the rate at which ambient DO is reduced, in part, because glycogen stores are finite [66][67][68] and, in part, because the accumulation of metabolic wastes from glycolysis may eventually reach toxic thresholds. For example, exceeding the plasma lactate threshold suggested for fish (~20 mmol L −1 ; [69,70]) seems to signal delayed mortality [71,72]. Indeed, some anoxia-tolerant ectotherms have evolved extraordinary strategies to mitigate the accumulation of these metabolic wastes. For example, goldfish (Carassius auratus) convert lactate to ethanol, which is a H + -consuming reaction [73,74] and the ethanol is produced easily and diffuses out of the fish across the gills provided the circulation is maintained. Thus, despite the drawback of losing carbon energy, ethanolproducing anoxic crucian carp (Carassius carassius) still maintain their routine cardiac activity [74] and do not become comatose [75]. In contrast, the anoxic freshwater turtle (Chrysemys picta) becomes comatose for months and avoids a fatal acidosis by demineralizing carbonates from its shell while at the same time complexing the lactate ions with calcium and magnesium [76].
Scope for hypoxic survival is necessarily time-limited and typically in the range of minutes for most fishes that do not tolerate severe hypoxia. Life below O 2crit and above ILOS also incurs an O 2 deficit, which can be measured with post-hypoxia O 2 uptake measurements [77]. Alternatively, the accumulated O 2 deficit (AOD) can be indirectly estimated from continuousṀO 2 measurements between O 2crit and ILOS if DO decreases at a steady rate. AOD is obtained by integrating the area bounded by the SMR threshold level and measuredṀO 2 values between O 2crit and ILOS [78]. This AOD calculation has many assumptions, all of which need to be experimentally checked. For example, an acceptable, standardized rate of hypoxia induction is needed. Also, the possibility of a metabolic suppression leading to AOD being an overestimate of the true O 2 deficit needs to be explored.
Metabolic suppression is certainly possible in hypoxia-acclimated fish. For example, European sea bass (Dicentrarchus labrax), an athletic teleost, suppressed SMR by 20% after hypoxic acclimation [51]. A more extreme metabolic suppression has evolved in anoxia-tolerant aquatic species (see review by [8]) like crucian carp (Carassius carassius) and freshwater turtles (Chrysemys picta bellii) which was estimated to reduce SMR up to 90% for up to 5 months [75,79]. Bushnell et al. (1984) speculated that hypoxia-acclimated rainbow trout (Oncorhynchus mykiss) suppressed SMR, but to reach this conclusion they extrapolatedṀO 2 measurements during swimming to a zero water velocity, a methodology that does not accurately measure SMR [80]. Hypoxia-acclimated steelhead trout (Oncorhynchus mykiss) maintained the same routineṀO 2 when measured in a normoxic thermal challenge [81]. Even so, a danger of using an RMR-derived O 2crit is that hypoxic suppression of behavioural activity could be mistaken as metabolic suppression. Likewise, using RMR rather than SMR to derive O 2crit could confound assessments of the time course of the onset of metabolic suppression because behavioural activity was initially suppressed rather than maintenance metabolism. Nonetheless, severe hypoxia or anoxia decreased RMR within~15 mins in tilapia (Oreochromis spp.) [62], <1 h in Pacific hagfish (Eptatretus stoutii) [77] and goldfish (Carassius auratus) [82] and <1.5 h in killifish (Fundulus heteroclitus) [83]. While the suppression of locomotor activities cannot be ruled out as a contributing factor, the large reductions in cardiac output seen during severe acute hypoxia in common carp (Cyprinus carpio), tilapia and freshwater turtles [62,84,85] suggest a rapid onset of metabolic suppression in some ectotherms, especially fishes. Therefore, the time course of SMR suppression will need further investigation as to its potential effect on AOD and even O 2crit determinations.
Clearly, O 2crit and ILOS determinations have unquestionably served fish biologists well for many decades. Yet, we predict measurement tools that identify sublethal effects of hypoxia will be more useful for conservationists and modellers of future hypoxic scenarios. To this end, Fry & Hart [10] provided a conceptual framework to holistically characterize the whole-organism function in a hypoxic environment using a hypoxia performance curve. Today, the hypoxia performance curve has taken the form of a limiting-O 2 concentration (LOC) performance curve for MMR, something we feel should be the most useful framework and assessment tool for an immediate addition to the toolbox for assessing hypoxic performance.

Individualizing the Limiting-Oxygen Concentration (LOC) Performance Curve
The underlying concept of a LOC curve for MMR is that maximum aerobic capacity (termed peak O 2 uptake,ṀO 2peak , for clarity if it is not measured in normoxia) must decrease once ambient DO produces a hypoxemic state ( Figure 2) and when an O 2 diffusion limitation exists at the gills. Consequently, asṀO 2peak decreases as a function of declining DO, AAS will decrease until O 2crit is reached and AAS is zero, as defined above. Making the requiredṀO 2 measurements to generate the LOC performance curve, however, is not a trivial task. First, there is the methodological challenge at a given DO is to assure a fish reaches a peak active state to obtainṀO 2peak . Then, as DO is reduced in a stepwise fashion, the peak active states need to be repeated sufficient times to accurately model the resulting curve that describes the decline inṀO 2peak with DO.
The first attempts to generate a LOC performance curve involved exercising a group of fish in a respirometer at one DO and then another group at a different DO [86,87]. This approach precluded the concern of a progressive depletion or build-up of metabolites if a fish or group of fish was repeatedly chased during progressive hypoxia. LOC curves derived in this fashion have been used in fishery models [52] and eco-physiological frameworks [53]. The disadvantage of this group approach is that the pooled data for different individuals, which generate the single curve, will smooth out the LOC performance curve and it cannot characterize individual variation, the very thing that natural selection acts on.
Rather than chasing groups of fish to generateṀO 2peak , fish can be swum individually in a swimming respirometer with a critical swimming speed (U crit ) protocol or a shorter ramped-U crit protocol. However, this methodological approach would be too timeconsuming to repeatedly measureṀO 2peak in this manner as a function of DO. First, the large water volume of the swimming respirometer takes a considerable time to equilibrate at each new DO level. Second, even a ramped-U crit protocol takes~1.5 h to complete. Thus, if say five DO levels were used to generate the LOC performance curve (likely the minimum number), it would take 7.5 h for completion without including the DO equilibration times. Consequently, swimming tests do not lend themselves well to an LOC curve determination.
MO 2peak is also measured by chasing individual fish to exhaustion outside a respirometer and then following post-exhaustion recovery of routineṀO 2 to measureṀO 2peak . An individual fish then could be exhausted at different DO levels before being placed into a respirometer at the same DO levels. This protocol could be used on different individuals at each DO level or on the same individual by repeatedly placing a fish in and out of the respirometer. The latter approach could be technically very challenging as well as unnecessarily stressful to the fish. Maintaining the chasing apparatus at the same DO as the respirometer would be another challenge. Of greater concern is that chasing a fish outside of a respirometer potentially underestimatesṀO 2max [88] and so this problem has been avoided by adding a simple chasing device to a standard respirometry chamber (Figure 3d). This way a fish can be agitated to generate a reliableṀO 2peak [88] and even repeatedly agitated without having to remove the fish from the respirometer. while it is continuously agitated for 15 min inside a respirometer (ṀO 2chase ) at normoxia (>80% air sat.). Grey vertical bars represent the flush period. PeakṀO 2 (ṀO 2peak , the horizontal broken line) is based on the average of the two highest values (two red dots as duplicates), which typically occur within the first 10 min, although a fish can reach or approach this peak several times. Group average ofṀO 2peak for 10 fish (c) shows that a commonality thatṀO 2peak is consistently elevated. MO 2peak values were estimated in a 1 min sequential interval regression algorithm. (d) A modified static respirometer to monitorṀO 2peak while fish is agitated to fatigue in the sealed respirometry chamber. The figure and data were adapted from [88].
This technical modification to a respirometer now means that DO inside the respirometer can be progressively reduced between repeated exercise bouts of an individual fish inside a respirometer [65,89]. Thus, it is possible to measureṀO 2peak as a function of declining ambient O 2 and generate an LOC performance curve for individual fish [51] and using far fewer fish than in a group-LOC protocol.
This individual LOC curve protocol will need further refinements to better conform with the 3-R principles (replacement, reduction & refinement) for animal care and for quality control purposes. For example, fish was exercised for 10 min first at a DO of 95% sat. DO was decreased progressively over 125 min,ṀO 2peak was repeatedly determined with a 5-min agitation period every 10 min at 10 additional DO levels [51]. Protocol improvement should consider the following.
First, the duration of the chase needs to be standardized. A 10-min chase may be unnecessarily long to generate aṀO 2peak , which is a key requirement. Crucial to refining this will be replacing the conventional, on-line sequential interval regression analysis oḟ MO 2 with an off-line analysis of the rolling regression algorithm coupled with a minimum sampling window (typically 1-2 min) [90]. This analytical protocol can better resolve the transient nature ofṀO 2peak by selecting the maximum and reliable (i.e., R 2 > 0.98) value from all the measurements recorded during and immediately (i.e., 5 mins) after the chase. Perhaps agitation could be stopped once two or three similar peaks were seen or pilot experiments could set a reliable period of chasing to ensure fatigue set in and 2-3ṀO 2peak values are recorded. Also, rather than mechanically agitating fish, it may be possible to use other stimuli to reliably generateṀO 2peak such as switch off of lights [91].
Second, the number ofṀO 2peak measurements per fish needs to be minimized to further minimize cumulative stress. The concern here is that the build-up of metabolites such as lactate and H + could limitṀO 2peak beyond that of DO alone. Likewise, if muscle glycogen stores become depleted, the fish will not be able to reachṀO 2peak . Both of these biological issues would manifest as a right-or down-shift of the LOC performance curve. Comparing fish tested at a single DO level with those tested individually at multiple DO levels could test for this possibility. Minimizing the DO levels tested on an individual, however, runs the risk of reducing the statistical power of accurately modelling the LOC performance curve. Obviously, the LOC curve is anchored byṀO 2max in normoxia (say > 90% sat.), but it does not have to be anchored by O 2crit . In fact, O 2crit might be better determined in a separate protocol for that fish and even before the LOC curve protocol starts to ensure that chasing a fish beforehand does not result in an overestimated SMR. A curve or breakpoint will likely require a minimum of five DO measurements for an accurate description. Preliminary experiments could be used to strategically place these DO levels on crucial parts of the curve. Ultimately the data will determine which is the best statistical approach to model these relationships, either a curvilinear fitting (e.g., one-phase association regression model; [40]) or a rectilinear fitting (e.g., segmented linear regression model; [35]), based on the inspection of the model quality (e.g., AIC values and R 2 ).
Third, a metric is needed to compare the LOC performance curves. In this regard, we feel the best comparative metric may be the DO that maintains 50% AAS (DO AAS-50 ), which borrows from the idea of P 50 of a blood O 2 equilibrium curve. If SMR andṀO 2max are known for each individual, eachṀO 2peak value can be expressed as a percentage of its absolute aerobic scope (AAS). Thus, an individual LOC performance curve can be fitted to 100% AAS [51], which adds sensitivity to the modelling by removing the inter-individual variation inṀO 2peak . The individual DO AAS-50 values can then be pooled to generate mean values for statistical comparison, say across species, at different temperatures for the same species and after acclimation to different hypoxic conditions that might represent future scenarios. Encouragingly, a 21% difference in DO AAS-50 was discovered for hypoxia-and normoxia-acclimated sea bass (38 vs. 48 % sat., respectively; p = 0.0031; Figure 4a; [51]). Moreover,ṀO 2peak at DO AAS-50 was not statistically different fromṀO 2max of sea bass acclimated to and tested at the same 50% DO (t-test: p = 0.065; Figure 4b; [51]).
In conclusion, we believe individual LOC performance curves hold promise for a broader application in physiology, ecology and evolution. An LOC performance curve can be generated in about three hours, with replication limited only by the number of individual respirometry chambers that are run simultaneously. Moreover, if DO is converted to a PO 2 , the PO 2AAS-50 can be directly compared with P crit and P 50 . Such comparisons could then provide novel mechanistic insights into hypoxic performance.

A Benefit of Characterizing Aerobic and Glycolytic Capacities on the Same Individuals
If respiratory phenotyping with maximum aerobic capacity, DO AAS-50 and O 2crit are performed on the same set of individuals, new analysis is possible beyond simple comparisons. One possibility is a regression analysis to ask if the various indices of aerobic and glycolytic capacities are correlated at the individual level, especially in athletic fish species. This possibility exists because most athletic fish species need a high aerobic capacity (i.e., oxidative phosphorylation) to sustain a high demand in locomotion, while short bouts of peak activity are greatly subsidized by glycolytic capacity (i.e., substratelevel phosphorylation). Some marine fishes even migrate vertically and diurnally between pelagic normoxic zones and coastal O 2 -minimum zones to actively forage in hypoxic habitats [92,93]. Yet, we are not aware of a comparative exploration to understand the correlations of whole-organism aerobic and glycolytic capacities in athletic fish species.
The advantage of an individual-based approach is that it considers the very variation that natural selection acts on. Furthermore, a standardized testing approach used in all the studies can greatly reduce the potential of including measurement discrepancies, which is possible when the broad comparison involves different test conditions. The Integrated Respiratory Assessment Protocol (IRAP), which measures 13 aerobic and anaerobic traits [78], has recently been used broadly to respiratory phenotype three athletic fish species in North America and Europe: rainbow trout in California and British Columbia, Atlantic salmon in Norway and British Columbia, and European sea bass in France.
One relationship that transcended all the species we examined was that O 2crit (difference: 1.8-6.1 folds, cv = 17.6-34.5%) and AAS (differences: 2.5-3.6 folds, cv = 20.6-31.1%) were quantitatively related at the individual level (y = −4.3x + 412.2, p < 0.0001, R 2 = 0.04, when all data were pooled). Thus, individuals that can better extract O 2 in hypoxia (a lower O 2crit ) also tended to have a higher aerobic capacity in normoxia ( Figure 5). Although significant correlations (p ≤ 0.042, 0.13 ≤ R 2 ≤ 0.63 for groups of significant relationships) did not necessarily exist for all testing temperature in every population that we considered (e.g., Californian rainbow trout strains tested at 12 • C), the relationships were remarkably consistent when a species was tested at the same temperature (p ≤ 0.0051, 0.22 ≤ R 2 ≤ 0.48). In fact, a warmer acclimation temperature tended to shift the regression lines to the right (Figure 5b,d for sea bass and rainbow trout). Acclimation to hypoxia will be one of the keys capabilities for fish species to succeed in a more hypoxic future. We know that fish have an incredible hypoxia acclimation capacity. For example, hypoxia acclimation increasesṀO 2max and aerobic scope in rainbow trout and Atlantic cod (Gadus morhua) [80,94], by improving O 2 uptake at the gills, O 2 delivered to the mitochondria and thereby the capacity to aerobically meet the metabolic needs [94,95]. The increased aerobic capacity in normoxia due to hypoxia acclimation likely attributes to the enhanced O 2 transport cascade as a beneficial carryover effect. As illustrated by the LOC curve, fish in hypoxia cannot fully sustain an aerobic capacity. Thus, an enhanced O 2 transport cascade likely contributes to an improved defence for aerobic capacity in moderate hypoxia. In this regard, hypoxia acclimation led to a leftshifted LOC curve and DO AAS-50 (Figure 4), showing fish can maintain the sameṀO 2peak to a lower DO, which should also translate to a better O 2 extraction in hypoxia (i.e., a decline in O 2crit ). Indeed, hypoxia acclimation leads to a decrease of O 2crit in epaulette sharks (Hemiscyllium ocellatum) [68], sailfin molly (Poecilia latipinna) [96], goldfish [97], Atlantic salmon (Salmo salar) [98] and snapper (Pagrus auratus) [99]. Still, future work is warranted to explore whether the phenotypic plasticity extends beyond the tolerance traits. In this regard, the LOC performance curve will advance a holistic characterization of the individual variations and plasticity for a hypoxia performance curve to quantify the sub-lethal effects. Although mortality events strip out the unfitted individuals, the tipping point that sets off a chain reaction of cascade events leading to mortality is the mechanisms of evolution at work, of which we have very little understanding.

Conclusions
To date, research on how fishes will live in a more hypoxic future mostly focused on tolerance traits rather than the sublethal effects at the moderate ambient O 2 levels. The reduced O 2 availability in the water suppressed the O 2 need for fish. Such idea is dispersed in the literature as a conceptual framework for a whole-organism hypoxic performance curve [10,11,36], experimental work of LOC performance curve for a fish population [53] and biochemistry concept of scope for survival [55]. By organically integrating these ideas as a tangible individual LOC performance curve, we will better understand the sublethal effects of a continuous shift of meeting the metabolic demands from sustained aerobic capacity to the time-limited anaerobic capacity as water O 2 content declines. Among these, a mid-point on the LOC performance curve (DO AAS-50 ) provides a comparative indicator for the individual differences in coping with the sublethal effects, where aerobic capacity is clearly suppressed, and substrate-level phosphorylation can be engaged. As the measurements of the hypoxia performance curve are carried out in high throughput aquatic respirometry systems (8, 16 and even 32 fish being run simultaneously), fish biologists are well poised to understand the standing individual variation of the hypoxic performance, one of the key pillars to better bridge physiology and ecology.