Introduction

Water temperature and oxygenation are critical factors that affect the performance of fishes, and both are being altered by human activities (Cahill et al., 2013; Barros et al., 2014; Gattuso et al., 2015; Luypaert et al., 2020; Albert et al., 2021). As such, it is not surprising that these two factors have been highlighted in a range of influential theories that attempt to use physiological parameters to understand the principles underlying the distribution and abundance of aquatic animals and how they may be affected by climate change. These include the theory of oxygen- and capacity-limited thermal tolerance (OCLTT) (Pörtner, 2002; Pörtner and Lannig, 2009; Pörtner et al., 2017), the oxygen- and temperature-limited metabolic niche framework (Ern, 2019), the theory of gill oxygen limitation (GOLT) (Pauly and Cheung, 2018; Pauly, 2021), and the concept of the metabolic index (Deutsch et al., 2015, 2020). While many of these theories remain controversial (Lefevre et al., 2017; Jutfelt et al., 2018), all emphasize the importance of interactions between oxygen and temperature. As such, there is an urgent need to understand the interactive impacts of warming waters combined with decreases in oxygen to predict how fish populations will cope with anthropogenic climate change.

Interacting Effects of Hypoxia and High Temperature

Temperature plays a critical role in determining water oxygenation through multiple mechanisms (Fig. 1). Increased temperature directly lowers water oxygenation by decreasing gas solubility, although this decrease may be largely offset by increases in the rate of diffusion as temperature rises, which increases oxygen delivery to the organism (Verberk et al., 2011). Climate change also indirectly affects water oxygenation by changing ocean circulation patterns and mixing, thus reducing the ventilation of the ocean (Keeling and Garcia, 2002; Matear and Hirst, 2003; Keeling et al., 2010; Somero et al., 2015). However, the largest effects of high temperature on aquatic oxygen levels occur because high temperatures increase the metabolic demand of a variety of microorganisms, such as bacteria and algae. This effect is particularly important in areas where human activities increase nutrient input into aquatic systems through runoff of agricultural fertilizer and wastewater, which reduces the nutritional constraint on the microbial growth (Smith and Schindler, 2009). These rapidly growing microorganisms can drastically decrease water oxygen levels and cause hypoxic dead zones in which fish cannot survive (Diaz and Rosenberg, 2008; Altieri and Gedan, 2015). This effect is exacerbated at night when photosynthetic organisms are respiring but not photosynthesizing (Diaz and Rosenberg, 2008; Diaz and Breitburg, 2009; Altieri and Gedan, 2015; Somero et al., 2015; Jenny et al., 2016). Because of these multiple interacting effects, aquatic hypoxia is likely to occur more frequently for longer periods of time, to more extreme levels, and across a wider range of habitats as our climate warms (Matear and Hirst, 2003; Diaz and Rosenberg, 2008; Diaz and Breitburg, 2009; Altieri and Gedan, 2015; Breitburg et al., 2018). For example, the mass mortality events known as summerkill that occur in lakes in the north-temperate zone, due to relatively transient episodes of high temperature and low oxygen, are predicted to increase more than fourfold by 2100 in northern latitudes (Till et al., 2019).

In addition to these environmental interactions, high temperature and hypoxia also have interacting effects on the fish themselves. These interactions are thought to be mediated through the joint impacts of high temperature and hypoxia on metabolism (Fig. 1; Fry and Hart, 1948; Precht, 1958; Prosser, 1969; Fry, 1971; Hazel and Prosser, 1974; Hochachka and Somero, 2001). Rising temperatures increase the rates of chemical and biochemical reactions (Schulte, 2015; Somero et al., 2017); and these thermodynamic effects result in increases in metabolic demand, which must be met with increases in metabolic energy supply for an organism to maintain energy balance. For many animals, this energy supply will be provided through aerobic metabolism, and this supply can become limited when environmental oxygen declines (McBryan et al., 2013; Schulte, 2015). These effects at the biochemical level cascade up to affect processes across levels of biological organization, with profound effects on complex physiological processes, such as cardiovascular function, muscle contraction, metabolism, energy budgets, and growth, which impact organismal growth and performance as well as thermal and hypoxia tolerance (Schulte, 2015; Somero et al., 2017; Little et al., 2020). The combination of high temperature and hypoxia in the environment is particularly devastating because high temperatures increase metabolic demand and simultaneously increase the prevalence and severity of environmental hypoxia, which limits the ability to supply this demand aerobically (Pörtner, 2001; McBryan et al., 2013; Deutsch et al., 2015; Schulte, 2015).

Timescales of Environmental Change

Both temperature and oxygen levels change in the environment across multiple timescales. Anthropogenic climate change is associated with a relatively gradual warming in temperature at a global scale, but it is also causing an increased frequency of heatwaves and extreme hypoxic events at local scales (Harvey et al., 2022). These two processes affect organisms across different timescales, and both are likely to be important in determining how well organisms will be able to cope with our changing world. The importance of mean temperatures for shaping the distribution and abundance of fishes is evident in the close relationship between the temperatures that maximize growth and reproductive performance in the wild and the mean temperature of warm range boundaries in fish (Payne et al., 2016). Similarly, the close correspondence of the metabolic index (a measure of temperature-dependent hypoxia tolerance) with biogeographic range boundaries supports an important role for mean temperature and oxygenation in shaping species distributions (Deutsch et al., 2015, 2020). However, there is substantial evidence that changes in the extent and frequency of extreme temperature events, such as heatwaves, are a major driver of biodiversity loss in the Anthropocene (Sandblom et al., 2014; Vasseur et al., 2014; Buckley and Huey, 2016; Frölicher et al., 2018; Román-Palacios and Wiens, 2020; Cheung et al., 2021; Harvey et al., 2022). The importance of extreme temperatures in shaping species’ distribution and abundance is clear in the close correspondence between laboratory measures of acute thermal tolerance, such as the critical thermal maximum (CTMax) and latitudinal patterns in thermal extremes (Sunday et al., 2019). Similarly, transient episodes of extreme hypoxia can have devastating effects on fish populations and likely shape fish species distribution and abundance. Indeed, a recent meta-analysis suggests that hypoxic events may be a more important driver of biodiversity change in the ocean than either temperature or ocean acidification (Sampaio et al., 2021). Thus, both long-term gradual changes in temperature and oxygenation and the increased frequency of extreme events pose challenges for aquatic organisms.

Physiological responses to, and tolerance of, short-term events, such as heatwaves and acute hypoxia, are likely to be quite different from responses to more gradual changes; thus, timescales of exposure to climate change-relevant stressors become critically important when attempting to predict the likely resilience or sensitivity of fishes to climate change. When acutely exposed to high temperature or hypoxia (such as during a heatwave or a hypoxic summerkill), organisms must rely on their existing tolerance mechanisms. In addition, tolerance limits can be modified by prior environmental experience through various forms of phenotypic plasticity acting at different timescales. Thus, organismal responses across longer timescales interact with their tolerance of acute events.

Types of Phenotypic Plasticity

Developing an understanding of the causes and consequences of phenotypic responses to hypoxia and high temperature is complicated by the fact that the literature on phenotypic plasticity is rife with alternative terms for similar processes or the same term being used in different ways (Piersma and Drent, 2003; Bowler, 2005; Ghalambor et al., 2015). In particular, physiologists and evolutionary biologists often approach this topic from very different perspectives and use different terminology (Huey and Berrigan, 1996); thus, it is important that we define the terms as we are using them here. We consider “phenotypic plasticity” to be a general term that describes the ability of an organism to change its phenotype in response to a change in the environment. In its broadest sense, this could refer to everything from instantaneous responses such as changes in heart rate to changes that persist over many generations. Whether very rapid and reversible phenotypic responses such as adjustments in heart rate and ventilation should be considered to represent phenotypic plasticity is a matter of debate (Woods and Harrison, 2002). Here we prefer to refer to these very rapid, and generally passive, effects occurring on a timescale of seconds to minutes as acute responses, although they have also been described as representing passive plasticity. We prefer to reserve the term “phenotypic plasticity” to refer to processes that involve active changes in biochemical properties that are regulated by the organism, such as changes in gene expression, membrane composition, or enzyme function. Here we discuss three main types of plasticity acting across different timescales (Fig. 2): acclimation, developmental plasticity, and intergenerational or transgenerational plasticity.

Acclimation is a type of plasticity that acts across relatively short time periods (hours, days, weeks, or months) and is generally considered to be reversible, such that the phenotype will revert to its original state if the environment returns to its original condition. Acclimation is thus often classified as an example of reversible plasticity, and the term “phenotypic flexibility” has also been proposed to distinguish this type of reversible phenotypic change from processes that result in more permanent changes in phenotype (Piersma and Drent, 2003; Woods, 2014). For simplicity, we also include phenomena such as heat hardening and the heat shock response with acclimation responses (Piersma and Drent, 2003; Bowler, 2005), because these processes are distinguished from the more classical view of acclimation and acclimatization largely based on the timescale and intensity of the exposure rather than on any fundamental difference in the process of the response (Bowler, 2005; Loeschcke and Sørensen, 2005).

Developmental plasticity occurs when the environment experienced during a particular critical window during development induces long-lasting, and often irreversible, changes in the phenotype (Burggren and Mueller, 2015). However, it is important to note that developmentally plastic traits exhibit a range of reversibility, depending on the type of exposure and the phenotype being examined (Burggren, 2020). There are also important interactions between developmental plasticity and acclimation (Beaman et al., 2016). For example, developmental exposures to environmental conditions such as altered temperatures can affect the capacity for acclimation at the adult stage (Scott and Johnston, 2012; Seebacher et al., 2014).

Intergenerational and transgenerational plasticity occur when the environment experienced by one generation affects the phenotype of individuals in subsequent generations (Mousseau and Fox, 1998; Jablonka and Raz, 2009; Ho and Burggren, 2012; Salinas and Munch, 2012; Donelson et al., 2018; Chang et al., 2021). We consider the term “intergenerational plasticity” to refer to changes acting across a single generation (from parent to offspring) that could be due to a wide range of maternal or paternal effects, including mechanisms such as transfer of lipids or hormones via the egg but that wash out in subsequent generations. We prefer to reserve the term transgenerational plasticity (which is also referred to as multi-generational plasticity) for phenomena that can be transmitted across multiple generations, usually meditated by more stable epigenetic mechanisms, such as changes in DNA methylation. However, we acknowledge that this distinction is not always made, and it can be difficult to draw the line between these phenomena because there are many potential interactions between them.

Is Plasticity Beneficial?

Plasticity of all types can result in significant phenotypic variation within and between populations, across life stages, and across generations and thus has the potential to play an important role in shaping species’ responses to climate change (Vagner et al., 2019). However, the extent to which plasticity of any type is adaptive (or even beneficial) has been the subject of extensive debate (Leroi et al., 1994; Wilson and Franklin, 2002; Woods and Harrison, 2002; Ghalambor et al., 2007; Seebacher et al., 2015). In general, the range of environments over which plasticity can be induced and the range over which the induced changes may be beneficial varies between organisms and between traits. Thus, the extent of beneficial plasticity of any given individual or species, if any, has clear limits (Auld et al., 2010; Norin and Metcalfe, 2019). Beyond some environmental threshold, a compensatory strategy may no longer be possible, and the organism may shift toward a passive endure-and-survive strategy. Ultimately, exposure to environments beyond those limits is likely to induce phenotypic changes that reflect harm or damage to the organism that could be beyond its capacity to repair.

The types of plastic responses that occur in response to environmental change have largely been considered in the context of acclimation. More than 65 years ago, Precht classified thermal acclimation responses into 5 groups (Fig. 3A), although these responses are relevant to plasticity acting at any timescale. This classification is based on the change in phenotype after acclimation relative to the effects of acute exposure to the stressor (Precht, 1958; Cossins and Bowler, 1987; Huey and Berrigan, 1996). For example, acute exposure to high temperature is expected to cause an increase in the rate of many biological rate processes (e.g., metabolic rate). After acclimation, if the phenotype returned to its control level, Precht (1958) considered this to be “complete compensation,” whereas “over-compensation” represented an overshoot in acclimation such that the rate after acclimation was lower than the control rate. Alternatively, in “inverse compensation” the rate following acclimation was greater than the rate with acute exposure. Note that Precht framed these responses in a context that implies that the ideal result of acclimation would be complete return to the phenotype observed under control conditions. However, as pointed out by multiple authors (Huey and Berrigan, 1996; Einum et al., 2019), for many traits (including metabolic rate) it is extremely difficult to establish, a priori, whether complete compensation is likely to be beneficial, because the optimal phenotype under the new environmental conditions is not always known. Thus, any one of these types of responses might represent beneficial acclimation or, alternatively, could be a pathological response. However, defining compensation for tolerance traits is relatively straightforward, because improved tolerance is unequivocally beneficial. Thus, complete compensation or overcompensation of tolerance would represent beneficial acclimation to high temperature, whereas inverse compensation would be consistent with a stressful or pathological effect.

The Challenge of Interacting Stressors

The majority of studies of resilience to environmental stressors have examined single stressors in isolation, whereas studies of the effects of interacting stressors are less common (Todgham and Stillman, 2013; Przeslawski et al., 2015; Gunderson et al., 2016; Jackson et al., 2016). In principle, when an organism is exposed to more than one stressor in combination, the resulting effect on performance may be additive, antagonistic, or synergistic (Folt et al., 1999; Crain et al., 2008; Todgham and Stillman, 2013; Gunderson et al., 2016; Fig. 3B).

For acute exposure to both high temperature and hypoxia in combination, the available evidence suggests that these stressors act synergistically on fish (Fig. 3B). Such exposure to high temperature and low oxygen together causes greater decrements in performance than would be expected based on the response to either stressor alone across a wide range of traits (Claireaux et al., 2000; Shimps et al., 2005; Schurmann and Steffensen, 2006; Nilsson et al., 2010; Healy and Schulte, 2012; McBryan et al., 2016; Townhill et al., 2017; Jung et al., 2020; Nudds et al., 2020; Somo et al., 2020). For example, acute transfer to high temperature markedly decreases the ability to tolerate acute exposure to hypoxia in most fish species (Fig. 3C).

Under control conditions at a moderate temperature, a fish will be able to maintain the appropriate orientation in the water column when exposed to severe hypoxia for a certain period (time to loss of equilibrium [LOE]), through a combination of increased oxygen extraction from the water and recruiting anaerobic pathways to help maintain adenosine triphosphate (ATP) supply. This time gets shorter when a fish is also acutely exposed to higher temperatures, because the higher temperatures cause an increase in metabolic demand, which causes a more extreme imbalance between energy demand and supply, shortening the time to LOE.

However, the exact extent of these interactions depends on the species and the particular levels of temperature and hypoxia that are tested (McBryan et al., 2016; Jung et al., 2020; Jensen and Benfey, 2022). The variability between species in the interactive effects of high temperature and hypoxia is clearly illustrated in a study by Jung et al. (2020), who examined effects on hypoxia tolerance at several different acute thermal exposures across multiple species. They found that some species exhibited negative effects on hypoxia tolerance with quite small increases in temperature (from 30 °C to 33 °C) and that other species did not show effects on hypoxia tolerance until temperatures increased from 30 °C to 37 °C (Jung et al., 2020). Following these acute effects, over time, depending on the species and the specific conditions, fish may induce plastic responses that alter their ability to tolerate hypoxia at high temperatures and that may change the time to LOE and result in one of the possible forms of compensation (Fig. 3C).

Most studies that have examined the nature and extent of various types of plasticity in response to exposure to high temperature or hypoxia have been conducted by examining the effects of each stressor in isolation. However, to understand and predict effects of human activities on aquatic biodiversity and to estimate the relative resilience of species as aquatic environments become warmer and more hypoxic, we must be able to untangle the independent and interactive effects of combined stressors and the extent of beneficial or harmful plastic responses to these combined stressors (Collins et al., 2021). Currently, the poor understanding of how fish will respond to changes in temperature and oxygenation makes it difficult to develop predictive models of how fish distribution and abundance will shift as our climate changes (Cahill et al., 2013; Todgham and Stillman, 2013). Therefore, our goal in the remainder of this review is to survey the current state of knowledge of the extent of plasticity in response to combined exposure to hypoxia and elevated temperature on fish across different timescales, with the goal of asking whether beneficial phenotypic plasticity has the potential to help fish cope with our rapidly changing environment.

Evidence for Beneficial Acclimation

Detecting acclimation in response to hypoxia or high temperature can be challenging because it requires information about the acute effects of the stressor. To illustrate this issue, consider the data presented in Figure 4A. In this experiment, Atlantic killifish (Fundulus heteroclitus (Linnaeus, 1776)) were acclimated to either 15 °C or 23 °C for several weeks, and then hypoxia tolerance was tested by measuring time to LOE at their respective acclimation temperatures (McBryan et al., 2016). Time to LOE decreased by 36% between fish acclimated at 15 °C and 23 °C, suggesting that high temperatures have a negative effect on hypoxia tolerance. In the context of climate change, these data suggest that fish will have more difficulty dealing with hypoxia as our climate warms. However, these data cannot be used to evaluate the extent, if any, of the acclimation response because the acute effects of exposure to high temperature are not accounted for. To assess the extent of the acclimation response, the data shown in Figure 4B are also needed. Here, hypoxia tolerance (time to LOE in severe hypoxia) was tested in fish acclimated to 15 °C after acute transfer to either 15 °C or 23 °C (McBryan et al., 2016). Acute transfer to 23 °C caused time to LOE to decline by 94% relative to fish tested at 15 °C. A comparison of the hypoxia tolerance of 15 °C acclimated fish acutely transferred to 23 °C (Fig. 4B) to the hypoxia tolerance of fish that have had time to acclimate (Fig. 4A) shows that there is considerable beneficial plasticity of hypoxia tolerance in response to thermal acclimation. However, fish acclimated and tested at 23 °C are still less tolerant of hypoxia than are fish acclimated and tested at 15 °C, indicating only partial compensation.

Only a subset of studies of acclimation to increased temperature and/or hypoxia have designs that are suitable to assess the extent of the acclimation response. For example, in a meta-analysis of studies examining plasticity in metabolic rate with temperature acclimation in ectotherms, Havird et al. (2020) showed that only 26% of studies across all ectotherms (or 30% in fish) contained information on the extent of both the acute effect of temperature and the phenotype after acclimation. Their analysis of the studies that control for these acute effects demonstrated that compensatory acclimation responses were common in fish and were particularly pronounced in marine species. However, the majority of studies showed only partial compensation of metabolic rate, and complete compensation was rare (Havird et al., 2020). There were also many cases in which no compensation was detected or where acclimation exacerbated the effect of acute exposure, resulting in inverse compensation. Similarly, variable results across taxa were detected using a somewhat different approach in a meta-analysis that calculated the temperature coefficient (Q₁₀) of metabolic rate and several related traits at lower levels of biological organization in both acutely exposed and thermally acclimated ectotherms (Seebacher et al., 2015). The results of these studies emphasize the importance of assessing both acute effects and the phenotype following acclimation if the goal is to assess the extent of plasticity. These analyses also clearly demonstrate that acclimation can only partially offset the acute effects of high temperature exposure in most species. However, it is also important to remember that the optimal phenotype under the new conditions is not always known, and complete compensation of the phenotype with acclimation may not always be the most beneficial response; interpretations of the benefits of patterns of partial compensation need to take this possibility into account.

We are not aware of any currently published meta-analyses examining the extent of the acclimation response to hypoxia in aquatic ectotherms that consider the acute effects of exposure. However, several meta-analyses have documented adverse effects of even modest hypoxia across a range of aquatic taxa, either acutely or in acclimated animals (Hrycik et al., 2017; Galic et al., 2019). Conversely, many studies suggest that fish have substantial capacity to compensate for the acute effects of hypoxia after acclimation (Lomholt and Johansen, 1979; Rees et al., 2001; Routley et al., 2002; Richards, 2009; Fu et al., 2011; Yang et al., 2013; Dan et al., 2014; Collins et al., 2016; Pan et al., 2017; Gilmore et al., 2019), suggesting that beneficial acclimation is possible in at least some taxa.

Studies of the interactive effects of high temperature and hypoxia on fish are relatively rare, and those that also account or control for acute effects that can be used to estimate the extent of acclimation responses are even scarcer. These studies have a variety of designs, which fall into three main categories: (1) studies of the effects of prior (usually brief) exposure to one stressor on the subsequent response to another stressor, (2) studies of the effects of acclimation to one stressor on the response to another stressor, and (3) studies of the effects of acclimation to both stressors simultaneously. The first two study designs allow the detection of a phenomenon known as cross-tolerance, in which the mechanisms that are induced to protect against one stressor also confer protection against a second stressor (Sinclair et al., 2013), whereas the third study design allows various types of interactions to be detected. Below we discuss the evidence for beneficial acclimation that has been obtained from each of these kinds of studies.

Evidence for Beneficial Developmental Plasticity

Studies of developmental plasticity can be performed in several ways. Some studies expose fish to altered conditions throughout the life cycle from development to adulthood, which mimics the likely effects of natural environmental exposures. That study design is highly relevant to understanding the resilience of fish to climate change but cannot distinguish between the effects of persistent developmental plasticity and acclimation. Instead, studies that expose fish to altered conditions during development (or even during specific critical times during development) and then return the fish to common conditions until testing at later life stages must be used to separate developmental plasticity from plasticity due to acclimation. A recent systematic review of studies of thermal developmental plasticity in CTMax across multiple taxa of ectotherms found that only 20% of the studies surveyed used designs that allow the detection of persistent developmental plasticity separate from effects of acclimation (Pottier et al., 2022). This suggests that even for studies of single stressors in isolation, more care needs to be taken with study design to establish whether developmental plasticity occurs and to separate it from the effects of acclimation.

Meta-analyses of studies that are able to detect developmental plasticity suggest that the direction and extent of developmental plasticity are highly variable between species and traits (Vagner et al., 2019). For example, temperature-induced developmental plasticity has been found to be beneficial, be deleterious, or have no effect on traits such as thermal tolerance (Chen et al., 2013; Moyano et al., 2017; Spinks et al., 2019; Illing et al., 2020), metabolic rate (Schaefer and Walters, 2010; Donelson et al., 2011; Veilleux et al., 2015), swimming performance (Batty et al., 1993; Johnston et al., 2001; Koumoundouros et al., 2009; Burt et al., 2012), and muscle phenotype (Koumoundouros et al., 2001; Scott and Johnston, 2012). Many of these differences are associated with small but persistent differences in gene expression (Scott and Johnston, 2012; Metzger and Schulte, 2018) that may involve differential expression of key regulatory genes involved in muscle development (Fernandes et al., 2006; Macqueen et al., 2008; Campos et al., 2013). Developmental temperature may also have effects on other behaviors related to swimming performance, such as foraging and predator avoidance (Vagner et al., 2019).

The pervasive but inconsistent effects of developmental plasticity in response to elevated temperatures are evident from a meta-analysis of published studies of developmental plasticity in thermal tolerance, in which the magnitude and direction of plastic effects on CTMax or LT₅₀ (the temperature lethal for 50% of the animals) were highly variable and on average did not differ significantly from zero (Pottier et al., 2022). However, one general pattern evident in these data is that the developmental stage at which the organisms are exposed appears to be important. For example, embryonic exposure to high temperature tends to have negative effects on thermal tolerance, while exposure at the juvenile stage tends to have positive effects (Pottier et al., 2022).

As is the case for developmental plasticity in response to altered temperatures, the effects of hypoxia during early development on adult phenotypes are quite variable (Vagner et al., 2019). Early developmental exposure to hypoxia has been shown to have long-term effects on growth (Johnston et al., 2013; Vanderplancke et al., 2015), sex ratios (Shang et al., 2006; Robertson et al., 2014), and, in some species, hypoxia tolerance (Wood et al., 2017). However, both beneficial and deleterious effects are observed, making it difficult to determine whether developmental plasticity in response to hypoxia is likely to provide any buffer against the effects of climate change.

The effects of developmental plasticity are thought to be mediated by a variety of epigenetic processes, although most studies to date have addressed the effects of temperature and have examined only changes in DNA methylation (Fernandes et al., 2006; Navarro-Martín, et al., 2011; Anastasiadi et al., 2017; Burgerhout et al., 2017; Metzger and Schulte, 2017). Similarly, changes in DNA methylation have been detected with early exposure to hypoxia (Grigaltchik et al., 2012; Crispo et al., 2020). Overall, the literature suggests that epigenetic changes may have lasting effects on gene expression of an organism and act as a mechanism underlying developmental plasticity in whole-organism phenotypes, but that these changes are not always beneficial.

One study, in zebrafish (Danio rerio (Hamilton, 1822)), has examined cross-tolerance between the effects of developmental temperature on hypoxia tolerance at later developmental stages (Levesque et al., 2019), and no evidence for developmental cross-tolerance was found between these stressors. Similarly, the literature examining the long-term effects of developmental exposure to hypoxia and high temperature in combination is very limited. However, the main general pattern that emerges is that there may be some beneficial effects of warm temperature development, but development in hypoxia often has negative effects. A fully factorial study of the effects of high temperature and hypoxia during development in European sea bass (Dicentrarchus labrax (Linnaeus, 1758)) revealed both reversible and non-reversible plasticity of traits (Cadiz et al., 2018). Hypoxia exposure had negative effects on body mass at the larval stage, although these differences did not persist once the animals were returned to control conditions at the juvenile stage. Similarly, developmental hypoxia had long-lasting negative effects on hypoxia tolerance (measured as O₂ at LOE), whereas developmental temperature had no effect on hypoxia tolerance (Cadiz et al., 2018). In an African cichlid (Pseudocrenilabrus multicolor (Schöller, 1903)), higher developmental temperatures caused a significant increase in CTMax in adults, but developmental hypoxia had no effect on adult thermal tolerance and there was no interaction between the factors (McDonnell et al., 2019).

An extensive experiment in Chinook salmon examined warm temperature and hypoxic exposure (50% dissolved oxygen) from fertilization to hatch, using a fully factorial design (Del Rio et al., 2021). Hatched larvae were then returned to cool, normoxic (control) conditions until the fry stage. At the alevin stage, routine metabolic rate was depressed to a similar extent by warm development, hypoxic development, and the combination of the two stressors; but these differences were not maintained at the later fry stage, demonstrating evidence of reversible plasticity. However, fry that had developed under warm conditions had greater hypoxia tolerance than fry that had developed under hypoxic conditions, suggesting a beneficial cross-tolerance in response to warm development and negative effects of development in hypoxia. However, whether these observed differences at the fry stage would persist to the adult stage through irreversible plasticity remains unknown and requires further investigation. Taken together, the data from the few available studies suggest that there are limited interactive effects of hypoxia and temperature through developmental plasticity and that the effects may be reversible for some traits. Indeed, many of the detected effects of developmental plasticity do not persist into adulthood. As well, interactive effects may not always be beneficial, and negative consequences for adult phenotypes are particularly evident following developmental exposure to hypoxia.

Evidence for Beneficial Transgenerational Plasticity

It has long been known that parental behavior and physiology can affect offspring phenotypes and performance, and both maternal (Green, 2008) and paternal (Crean and Bonduriansky, 2014) effects have been widely documented. It has generally been assumed that maternally derived factors, including materials such as lipids, hormones, and RNA, drive intergenerational effects, based on the relative contribution of the sperm and egg to the zygote (Mousseau and Fox, 1998). However, recent studies (Jiang et al., 2013) have shown that paternal methylation patterns can be stably inherited, suggesting that maternal factors are not alone in influencing offspring phenotype and that sperm can transfer information to embryos via a variety of routes (Curley et al., 2011; Kumar et al., 2013; Crean and Bonduriansky, 2014; Donkin and Barrès, 2018). In general, both intergenerational and transgenerational effects can also be mediated through epigenetic mechanism, such as changes in DNA methylation and histone acetylation. All these mechanisms represent forms of non-genetic inheritance, in which offspring inherit something other than just the DNA sequence from their parents. Many forms of non-genetic inheritance, such as those involving transfer of substances such as lipids in the oocyte, are primarily expected to influence only the next generation, with impacts being expected to dissipate across subsequent generations. Conversely, epigenetic mechanisms, such as DNA methylation, at least have the potential to be stably inherited across generations and, thus, could mediate transgenerational plasticity (Mousseau and Fox, 1998; Badyaev and Uller, 2009; Metzger and Schulte, 2016).

Transgenerational plasticity has been detected in a diverse range of taxa from unicellular eukaryotes (Nowacki et al., 2011) to mammals (Blewitt et al., 2006), but the existence and adaptive significance of phenotypic plasticity mediated by transgenerational inheritance is controversial (Horsthemke, 2018; Perez and Lehner, 2019). The challenge of detecting transgenerational plasticity, particularly when the mother is exposed to the environmental stressor, can be illustrated by considering the reproductive biology of vertebrates (Kovalchuk, 2012). In mammals, exposure of a pregnant female exposes both the F1 (embryo) and F2 (embryonic germ cells) generations to the initial stressor (Skinner, 2008). As a result, demonstrating transgenerational plasticity in cases where the mother is exposed requires characterization of phenotypes in at least the F3 generation, because direct exposure of the embryo cannot be ruled out as the cause of the plasticity. The situation is somewhat more straightforward in fish; however, even in fish, germline cells (F1) develop within the gonads of the F0 generation several hours after fertilization (Kurokawa et al., 2006) and, thus, maternal exposure to altered environments may also result in exposures of the F1 germline. Thus, it is necessary for the phenotype to persist to the F2 generation to detect transgenerational plasticity (Mousseau and Fox, 1998; Jablonka and Raz, 2009; Ho and Burggren, 2010). The need for studies across multiple generations has resulted in the majority of transgenerational plasticity studies in fish using small-bodied fish with short generation times (e.g., sheepshead minnow Cyprinodon variegatus Lacepède, 1803) (Salinas and Munch, 2012) and zebrafish (Knecht et al., 2017)) as model organisms; thus, little is known about this type of plasticity in other species of fish.

There are multiple studies that suggest that plasticity across a single generation occurs in a wide range of fish taxa, including those in freshwater (Ho and Burggren, 2010, 2012; Lee et al., 2020; Chang et al., 2021; Penney et al., 2021), estuarine (Salinas and Munch, 2012; Chang et al., 2021; Munch et al., 2021), and marine (Donelson et al., 2012; Shama and Wegner, 2014; Veilleux et al., 2015; Shama et al., 2016; Ryu et al., 2020) environments (for review see Munday, 2014; Donelson et al., 2018). For example, in fish, intergenerational plasticity has been detected in response to parental temperature for aerobic scope (Donelson et al., 2012), growth (Salinas and Munch, 2012; Chang et al., 2021; Munch et al., 2021), body size (Shama and Wegner, 2014), sex ratio (Pierron et al., 2021), and hypoxia tolerance in response to parental exposure to hypoxia (Ho and Burggren, 2012; Ragsdale et al., 2020). Studies across multiple generations (demonstrating transgenerational plasticity) are rarer, but beneficial effects of elevated grandparental temperature on juvenile growth across two generations have been detected in sheepshead minnows (C. variegatus), threespine stickleback (Gasterosteus aculeatus Linnaeus, 1758), and spiny chromis damselfish (Acanthochromis polyacanthus (Bleeker, 1855)) (Shama and Wegner, 2014; Donelson et al., 2016; Lee et al., 2020). In contrast, hypoxia exposure has been shown to have negative effects on reproductive fitness across several generations (Wang et al., 2016; Lai et al., 2019). For example, hypoxia causes female marine medaka (Oryzias melastigma (McClelland, 1839)) to alter DNA methylation and gene expression patterns and is associated with a dramatic decrease in hatching success through to the F2 generation (Lai et al., 2019).

A meta-analysis of transgenerational plasticity across both plants and animals for a variety of environmental exposures and phenotypes suggests that the mean effect size for transgenerational plasticity is not distinguishable from zero, with both positive and negative plasticity being detected, depending on the study (Uller et al., 2013). This pattern suggests that beneficial transgenerational plasticity is not a general phenomenon and may be present only in a subset of species or under very specific conditions (Kovalchuk, 2012). For example, spiny chromis damselfish demonstrate beneficial transgenerational plasticity in aerobic scope (a proxy for performance and fitness) with warming of +1.5 °C, whereas fish warmed at +3.0 °C displayed reduced aerobic scope (Donelson et al., 2016). These observations are consistent with the idea that high-intensity stress is less likely to lead to adaptive plasticity compared to familiar, moderate-intensity stress (Parsons, 1994, 2005). The total length of exposure can also strongly influence whether beneficial transgenerational plasticity is detected. For example, sheepshead minnows required at least 30 days of exposure to elevated temperatures (+10 °C) to show beneficial transgenerational plasticity, whereas shorter parental exposures had no effect (Salinas and Munch, 2012).

The adaptive significance of transgenerational or intergenerational plasticity, if present, is likely to be dependent on the extent to which the environmental conditions of the parental generation predict the environment that will be experienced by future generations. This idea has strong empirical support from studies in sheepshead minnows (Munch et al., 2021). In this species, individuals from less predictable environments exhibited less plasticity across generations than individuals from more predictable environments (Munch et al., 2021). These observations are consistent with the hypothesis that transgenerational plasticity is generated only under a very particular set of circumstances that are likely to vary across species.

There are very few, if any, studies of transgenerational effects of combinations of hypoxia and temperature (Donelson et al., 2018). One study has examined the potential for transgenerational cross-tolerance between these stressors and found that exposing the parental generation of zebrafish to hypoxia leads to increased hypoxia tolerance in offspring but does not significantly affect thermal tolerance, indicating an absence of transgenerational cross-tolerance (Ho and Burggren, 2012). More studies of combined stressor exposures and cross-tolerance are needed to allow assessment of the potential for transgenerational plasticity to act to increase the resilience of fishes to climate change.

Conclusions and Implications

Although studies of the extent of plasticity in fishes in response to exposure to the combination of high temperature and hypoxia are relatively rare, it is clear that most fish have only limited capacity to compensate for the synergistic negative effects of acute exposure to these stressors. In particular, cross-tolerance responses appear to occur under only a relatively narrow range of circumstances and only in a few species. Thus, this type of within-generation plasticity is likely to have only limited utility for buffering fishes in the face of climate change, particularly at species’ range edges, where available plastic responses may already be recruited to cope with current conditions. However, our ability to make strong conclusions about the likely benefit of acclimation responses is limited by the relatively small number of studies in which both acute and acclimated phenotypes have been assessed, which makes assessing the extent of phenotypic plasticity impossible. The limitations on plasticity are hinted at, however, by comparisons of animals acclimated to future conditions that suggest the presence of ongoing negative impacts of exposure to high temperature and hypoxia that cannot be fully compensated.

Developmental plasticity seems even more unlikely to provide significant compensatory effects. In general, there is some evidence for beneficial developmental plasticity in response to either hypoxia or temperature alone, although hypoxic exposures during development also often have deleterious effects. Combined exposures have rather variable effects, depending on the species and the exposure conditions; and the detrimental effects of hypoxic exposure may undermine any beneficial effects of development at warmer temperatures. To properly assess the extent of developmental plasticity and distinguish these effects from the effects of acclimation, it is necessary to expose fish to altered conditions during development and then return them to control conditions until phenotypes are assessed at subsequent life stages. However, in the context of climate change, fish are much more likely to be exposed to altered conditions during both development and subsequent life stages. There are very few studies that have examined these types of life-cycle exposures to assess the possibility of combinations of developmental plasticity and acclimation to increase resilience in the face of climate change. The intergenerational and transgenerational effects of exposure to combinations of hypoxia and high temperature are largely unknown, and studies of this sort are desperately needed to understand the likely impacts of climate change on fishes.

Although this assessment paints a rather bleak picture of the potential for phenotypic plasticity (at any timescale) to fully protect fish populations from either the gradual increases in temperature and hypoxia or the increased frequency of extreme events, there is also substantial between-individual variation in plasticity in fish (Seebacher and Little, 2021), as well as interindividual variation in tolerance of both high temperature and hypoxia (Strowbridge et al., 2021). Thus, it is possible that our changing climate will impose selection on this standing genetic variation that may provide some resilience of fish populations in the face of combined exposures to high temperature and hypoxia as our climate changes.