Global change and physiological challenges for fish of the Amazon today and in the near future

Amazonia covers 7.76 million square kilometers, more than 5 million of which are in Brazil. The area encompasses rainforests (78%), Andean areas (1.5%), flooded areas (floodplains and igapós) (5.83%), savannas (12.75%) and two tropical steppes (1.89%) (Hilty, 2012). Between these macro-zones, there are transition areas (ecotones) with mixed characteristics, such as the Amazon-Cerrado and the Amazon-Pantanal. Within the rainforest itself, there are 53 distinct environments, each with unique temperature and precipitation patterns that impose distinct selective pressures on resident organisms (Adeney et al., 2016; Hess et al., 2015). The biome is home to the longest river valley on Earth: the Amazon River (6992 km) arises in Peru and discharges 250,000 m³ s⁻¹ into the Atlantic at Cabo do Norte, representing 20% of all of the freshwater entering the world's oceans (Sioli, 1984; Val et al., 2006; Val and Almeida-Val, 1995). Its tributaries include three more of the world's longest rivers (Rio Negro, Rio Madeira and Rio Japurá), and 20 more of the 34 largest tropical rivers (Latrubesse et al., 2005). These waters drain from diverse geologies, and the final discharge represents a mixture of three distinct water types characterized by their colour (Sioli, 1984) (Fig. 1). This is a result of 65 million years of geological rearrangements that include tectonism and changes in climate operating in a very singular environment.

All Amazon waters are ion-poor, but those originating from the Andes (‘white waters’) with a muddy, cafe au lait colour have the most ions and a pH close to neutrality. White waters carry suspended particulates from the mountains, down the Rio Solimões and its tributaries, all the way to the Atlantic mouth of the Amazon River. In contrast, the ‘black waters’ of the Rio Negro watershed are darkly coloured by their high content of dissolved organic carbon (DOC) (see Glossary) and have the lowest ion levels and pH (2.5–6.0), reflecting the mineral-poor soils of their jungle origin. The black (Rio Negro) and white waters (Rio Solimões) merge just east of Manaus. The third major water type, the ‘clear waters’, are also very dilute but low in DOC and particles, with a higher pH (6.0–6.8). These drain from ancient, eroded surfaces of the Guiana massif and central Brazil, the headwaters of certain Andean rivers, and clear-water streams of the central lowlands. These three different environments pose immense challenges to endemic fish species, some of them transiting between different water types, such as the freshwater sardine (Araújo et al., 2017).

The most important seasonal variation throughout the region is the annual hydrological cycle or ‘flood pulse’ (see Glossary), where the water levels vary dramatically throughout the year and rhythmically over the years (Junk et al., 1989; Schöngart and Junk, 2007) (Fig. 2). These flood and drought dynamics are the main ecological driving force shaping evolutionary and adaptive processes in Amazonia (Kramer et al., 1978; Sioli, 1984; Junk et al., 1989; Assahira et al., 2017). Water levels serve as a proxy for the profound environmental variations that demand continuous physiological adjustments by resident organisms (Kramer et al., 1978; Val and Almeida-Val, 1995). Rising waters reconnect water bodies, facilitating seasonal interchange of fish fauna (Val and Almeida-Val, 1995). Periodic inundation and desiccation of plains and rainforests change habitat size, opening up and removing feeding, breeding and migration opportunities for the fish (Cox-Fernandes, 1997; Moura and Val, 2019). It also mobilizes more or less DOC, particles and nutrients, and creates cycling CO₂, pH and O₂ regimes through the respiration and photosynthesis of algae and vegetation (Val and Almeida-Val, 1995; Assis et al., 2019). Variations in O₂, which may range from virtual anoxia to several-fold supersaturation (hyperoxia), occur spatially, daily and seasonally, and are particularly important (Val and Almeida-Val, 1995).

This Review focuses on Amazon fish under climate change scenarios. Much remains to be known, but what we do know allows us to assess the potential effects of these changes. To this end, we present a rapid review of the biological diversity of Amazonian fish, the status and predictions of climate change in the Amazon, and what we know about the physiology of Amazonian fish that may be influenced by climate change (altered pH, DOC, suspended particles, temperature, P_CO2 and P_O2 levels). Finally, based on what we know, we suggest five classes of effects of global changes on Amazon fish.

The Amazon basin hosts the richest ichthyofauna in the world – more than 2700 species representing approximately 15% of all freshwater species (Jézéquel et al., 2020; Tencatt et al., 2020). New species are still being discovered (Tencatt et al., 2020) or added by range extension, for example via the Casiquiare corridor to the Oronoco basin (Winemiller et al., 2008). Species diversification processes are still occurring (Albert et al., 2018; Dutra et al., 2020; Sá-Leitão et al., 2021). Other aquatic groups exhibit similar ongoing speciation, demonstrating the importance of Amazonia as a hotbed of neotropical biodiversity (Antonelli et al., 2018).

The diversity of fish encompasses an array of physiological and morphological specializations to cope with environmental challenges. Table 1 presents a list of the major families of the Neogene fish fauna of Amazonia (Lundberg et al., 2010), with highlights of each group's physiology. This immense Amazonian ichthyofauna includes representatives from the most basal teleosts (e.g. Osteoglossiformes) to the most specialized groups (e.g. Cichliformes, until recently included among the Perciformes). Many peculiar species with unique characteristics are found here, such as the only stenohaline (see Glossary) freshwater elasmobranchs, the stingrays (Myliobatiformes, formerly Rajiformes) of the Rio Negro (Wood et al., 2003) and the giant osteoglossid pirarucu, among the largest freshwater fish in the world, with its red blood cells containing the potent haemoglobin allosteric effector inositol pentaphosphate (IPP) (Bartlett, 1978; Isaacks et al., 1977; Val et al., 1992). Amazonia also hosts an immense variety of catfish (Siluriformes), some of them armored, such as the acari-bodo that uses the stomach-intestine as an accessory organ for respiration (Cruz et al., 2008; Val et al., 1990), and others of large size, especially the piraiba (Petrere et al., 2004). There are also various electric fish (Gymnotiformes), notably the poraquê (electric eel), which can discharge up to 600 V (Catania, 2014; Val and Almeida-Val, 1995), and various serrasalmids (Characiformes), such as the legendary piranhas, and the tambaqui, which can expand its lower lips to breathe at the water surface (Saint-Paul, 1984; Val and Almeida-Val, 1995; Val and Oliveira, 2021). Some Amazonian fish species are already living at the limits of their biological abilities, such as the small characids inhabiting small forest streams. Thus, the rapid environmental transformations worsened by climate change can be devastating for native fish.

CO₂ in the Earth's atmosphere has risen from approximately 280 ppm in 1850, when the industrial revolution was taking place mainly in Europe, to more than 411 ppm in 2019 (de la Vega et al., 2020). The greenhouse effect (which results also from emissions of CH₄, N₂O and other anthropogenic gases), is compounded by other anthropogenic activities such as deforestation, wildfires, road and dam construction, and water harvesting and diversion. The Earth's surface has warmed by approximately 0.8°C in the last 40 years alone (Gross, 2018), with 2016 being the warmest year (+0.9°C) since 1950 (Marengo et al., 2018). Consequences include hotter and more widely fluctuating water temperatures, acidification and salinization of water bodies, increased mobilization of DOC, changes in DOC chemistry and decreases in dissolved oxygen (Lloyd and Shepherd, 2020; Scanes et al., 2020).

Amazonia is particularly vulnerable to the ecological impacts of predicted climate change (Castello and Macedo, 2016) and will face less rainfall, warmer and drier periods, and more extreme events, including more intense droughts and floods (Brodie et al., 2012; Dai, 2012), which could lead to the disappearance of the forest in eastern, southern and central Amazonia. This is being compounded by direct deforestation; 12% of the rainforest has already been lost to logging and agriculture (Stropp et al., 2020).

Decreasing deforestation rates in the earlier part of the 21st century were an encouraging trend (Nobre et al., 2016), but this has now been reversed owing to changes in government policy. Deforestation rates are now the highest of the past decade (Jaffé et al., 2021). At a threshold somewhere between 20 and 40% forest loss, often called a tipping point, it will be nearly impossible to recover the system: forest will be permanently replaced by savannah, and temperatures will further increase (Lovejoy and Nobre, 2018). These changes will reduce the carbon-removal capacity because photosynthetic CO₂ fixation will decrease, while more decaying organic matter will be available for respiratory CO₂ production. Therefore, net CO₂ emissions will rise, further exacerbating the problem. Additionally, as forest land is converted to agriculture and savannah, major nutrients and ions are released into the water through the process of salinization (see ‘Increased ion levels’, below; Biggs et al. 2004; Chaussê et al. 2016; Markewitz et al. 2001). There is a clear connection between vegetation cover and moisture recycling, which in turn relates to the water, energy and carbon cycle in the region (Marengo et al., 2018).

Given the geographic and geological heterogeneity of Amazonia, climate change will differentially affect its various parts, with a mosaic of complex effects. If uncontrolled and unsustainable land exploitation continues, temperature projections will far exceed those experienced to date (∼1°C). Predictions for the period 2041–2070 indicate that the northeastern region of the state of Amazonas could experience the greatest warming, with a further increase of up to 5°C. In contrast, the central region of the state of Amazonas and the upper Rio Negro will encounter the smallest thermal changes, with an increase of up to 3.5°C (Menezes et al., 2018). Regardless, both temperature elevations will have devastating effects on native fish. Unlike terrestrial and marine organisms (Jacox et al., 2020; Lenoir et al., 2008), there is little opportunity for Amazon fish to move up mountains or towards the poles to escape extreme temperatures and associated environmental changes. Some models indicate that one-third of the fish species in Amazonia, as well as in the neighbouring Tocantins and Araguaia watersheds, may face extinction if these changes persist in conjunction with the most extreme climate change scenarios (Xenopoulos et al., 2005).

Although the overall trend for Amazonia as a whole is desiccation, increases in precipitation and increased discharges of Amazon River tributaries are currently occurring in the Northern Andean region (Pabón-Caicedo et al., 2020). More intense and more frequent floods are now occurring, with a 5-fold increase after the year 2000, compared with the first half of the 20th century (Barichivich et al., 2018). There are two related implications. The first is that the amplitude between flood and ebb has increased significantly over the last 25 years (see Fig. 2) (Schöngart and Junk, 2020). Therefore, during extreme floods the fish have access to a greater extent of flooded forest, with impacts on their natural history, particularly regarding refuge, dispersal, feeding and reproduction. The second is that increased discharge from tributaries puts a greater amount of suspended sediments into the central channel of the Amazon River, and will also elute more DOC from the submerged jungle soils and vegetation. Both higher suspended particles and higher DOC levels will have important effects on fish physiology. Estimates for the end of the 21st century suggest that the Andean region may become even warmer (Pabón-Caicedo et al., 2020). This additional warming may be locally amplified because of the reduction of vegetation cover, in a spiralling effect, because many tropical plant species will not germinate at higher temperatures (Sentinella et al., 2020), thereby contributing to environmental imbalances in both this region and the entire Amazon basin.

As temperatures rise and flooding events become more intense, hypoxic events are also becoming more intense and of longer duration, especially in floodplains and igapós (A.L.V., personal observation). Not only is the solubility of O₂ reduced, but the metabolic demand for O₂ is increased. Therefore, depletion of dissolved O₂ and build-up of dissolved CO₂ owing to respiration of greater amounts of organic matter happen more quickly, at a time when the fish's metabolism needs more O₂. Associated increases in DOC and increases in P_CO2 will further compound the problem by lowering pH in the poorly buffered Amazonian waters.

As previously discussed, in the future, Amazon waters will be more acidic, darker (i.e. more DOC and more suspended particles), hotter, higher in ions, higher in CO₂ and lower in O₂. We will now use these forecasts to predict the resulting physiological effects on the native fishes of Amazonia.

Amazon waters are extremely dilute, low in alkalinity, poorly buffered and, therefore, easily acidified. However, many Amazon fishes, especially those native to the black waters of the upper Rio Negro watershed, are already adapted to acidic environments. Here, dissolved organic acids (humic substances) contribute approximately 85% of the acidity, and the high P_CO2 levels contribute only approximately 15% (Leenheer, 1980). Waters with a pH ≤4.5 will suffer little further acidification, considering the local geochemical characteristics and the logarithmic nature of the pH scale, so the devastation seen with the acid rain crisis of the 1970s–1990s in many areas of the Northern Hemisphere (Schindler, 1988; Wood, 1989) will likely not occur in the black-water ecosystems. However, in the less acidic lower Rio Negro, as well as in white waters, clear waters and mixing zones, increases in environmental P_CO2 and acidic precipitation may lower pH in waters currently in the pH 4.5−7.0 range, and thereby impact the organisms adapted to live there.

From studies on temperate species, we have learned that the damaging physiological effects of low pH are on ionoregulation (see Glossary), rather than on acid–base balance (see Glossary), as long as water calcium concentrations are relatively low (<10 mg l⁻¹=<250 µmol l⁻¹, ‘hardness’ <25 mg l⁻¹ as CaCO₃ equivalents; Wood, 1989). Virtually all Amazon waters fit this description, with calcium concentrations below 10 mg l⁻¹ (Val and Almeida-Val, 1995). The responses of native fish to low pH seem to fit this pattern, with little effect on acid–base regulation or the related function of ammonia excretion (Wilson et al., 1999; Wood et al., 2018, 1998). In contrast, many studies have documented disturbances of ionoregulation, as ionoregulatory capacity is correlated with realized niche (see Zimmer et al., 2021). Nevertheless, Amazon fish are generally quite tolerant of low pH, with thresholds for serious effects being 1–3 pH units lower than in temperate species, though with considerable interspecific variation (e.g. Duarte et al., 2013; Gonzalez et al., 1998, 2002; Wilson et al., 1999). The disturbance invariably involves elevated rates of net Na⁺ and Cl⁻ loss at the gills. These losses provoke a fluid shift out of the blood plasma into the intracellular compartment and an accompanying concentration of red blood cells (Wood et al., 1998). The fish may eventually die of cardiovascular failure associated with this haemoconcentration (Milligan and Wood, 1982).

Most studies on low pH effects have focused on Na⁺ balance at the gills. The effects are two-fold: (i) inhibition of active Na⁺ uptake and (ii) stimulation of passive Na⁺ efflux (outflux), with the latter generally making the larger contribution to net Na⁺ loss. Using classic Michaelis–Menten kinetic analysis (see Glossary) and pharmacological tools, Gonzalez, Val and colleagues have identified two basic patterns of ionoregulation in fish adapted to these acidic, ion-poor waters (Duarte et al., 2013; Gonzalez et al., 2006, 1998, 2018, 2020, 2021, 2002; Gonzalez and Preest, 1999; Gonzalez and Wilson, 2001; Preest et al., 2005; Wood et al., 2003, 2002, 2014). There may be separate phylogenetic origins for these two strategies (Gonzalez et al., 2017, 2018, 2020). However, to date, only three of the 18 piscine orders endemic to the Amazon have been characterized as having low pH responses, so there may well be other response patterns.

In the first (Fig. 3A), typical of the Cichliformes, as well as freshwater stingrays (Myliobatiformes) endemic to the Amazon, the Na⁺ uptake system has a relatively low affinity (i.e. high K_m), whereas transport capacity (J_max) is variable. Both unidirectional influx and efflux rates are low but in approximate balance at moderate pH values. Na⁺ uptake is strongly inhibited by low pH (3.0–4.5), and by standard pharmacological blockers commonly used to identify the Na⁺/H⁺ exchange or coupled Na⁺ channel/H⁺ ATPase systems seen in most other fish, suggesting the presence of the same Na⁺ transporters as well documented in model species such as trout and zebrafish (Clifford et al., 2021; Dymowska et al., 2012). However, branchial efflux is extremely resistant to stimulation by low pH, so net loss rates are not high, and can probably be compensated by uptake from the diet. These species have a ‘wait it out’ strategy and may be the ones more affected by future acidification.

In the second (Fig. 3B), as seen in the Characiformes, the Na⁺ uptake system has a very high affinity (i.e. low K_m) and high capacity (i.e. high J_max), so that at moderate pH values, unidirectional uptake rates are high, keeping balance with comparably high rates of efflux. The Na⁺ uptake system(s) are unusual and as yet poorly understood, being extremely resistant to inhibition by low pH (3.0–4.5) or any of the standard Na⁺-transport blockers commonly effective in most other fish (Clifford et al., 2021). The diffusive efflux rates across the gills, while high, are again very resistant to the stimulatory effect of low pH. These fish have a ‘business as usual’ strategy and may be the ones more resistant to future acidification.

In temperate species, increased branchial efflux rates of major ions appear to be caused by the leaching of Ca²⁺ away from the tight junctions between branchial epithelial cells (McDonald et al., 1980; McWilliams, 1982), accompanied by a shift to less negative/more positive values of the transepithelial potential (TEP) across the gills (reviewed by Gonzalez et al., 2006). As the TEP at the freshwater gill is largely a diffusion potential (Potts, 1984), this suggests that the Cl⁻/Na⁺ permeability ratio also increases. In several Amazon fish, 10-fold elevations in water calcium protected against the elevation of Na⁺ efflux at low pH (Gonzalez et al., 1998; Wood et al., 2003, 1998) and also against the rise in TEP (Wood et al., 1998). However, in several other black-water species, there was little or no protection provided by calcium (Gonzalez et al., 1998; Gonzalez and Preest, 1999). Other mechanisms may be involved, such as upregulation of tight junction genes in the gills, as reported for claudins in the sardine, a characiform that inhabits white, black and clear waters (Araújo et al., 2017). As discussed subsequently, the DOC in the black waters may also play a key role.

Increases in major ion levels are directly associated with deforestation and agriculturalization (Biggs et al., 2004; Chaussê et al., 2016; Markewitz et al., 2001). Overall, these may be beneficial in helping the uptake of Na⁺, Cl⁻ and Ca²⁺ by fish in the ion-poor waters, by shifting the uptake and balance points on the Michaelis–Menten relationships (Fig. 3) to a higher level, and possibly also by lowering the efflux rates by the ‘tightening’ action of Ca²⁺ on gill permeability. However, they may change species assemblages and facilitate invasions by exotic species less tolerant of ion-poor waters.

DOC encompasses a variety of large, heterogeneous molecules originating from the physical breakdown or microbial processing of formerly living materials (Leenheer, 1980; Thurman, 1985). Carbon makes up approximately 50% by mass of these molecules. DOC concentrations as high as 30 mg l⁻¹ have been recorded, but more typical levels are 5–15 mg l⁻¹ in black waters, 2–5 mg l⁻¹ in white waters and 1–5 mg l⁻¹ in clear waters. Amazon DOCs are chemically diverse among the three water types (Gonsior et al., 2016; Leenheer, 1980), but in all, allochthonous (see Glossary) sources dominate – i.e. of terrigenous origin, derived largely from plant molecules such as lignins. Humic substances are defined as those organic compounds that cannot be classified as any other chemical class of compounds (e.g. polysaccharides, proteins, etc.) (Gaffney et al., 1996). They are often highly coloured owing to the presence of aromatic rings (phenolic groups) and constitute >50% of most Amazon DOCs. A variety of smaller, less coloured, more hydrophilic molecules (proteins, amino acids, carbohydrates, fatty acids, small organic acids, etc.) comprise the remainder. However, in black waters, violacein, a purple pigment with anticancer, antimicrobial and antiparasitic properties produced by the abundant microbe Chromobacterium violaceum, contributes colour to the latter fraction (Caldas, 1990; Duran et al., 2007). Its effects on fish remain unknown. Within the humic substances, the proportions of humic acids (higher molecular weight, defined by precipitation at pH 1.0–2.0; Gaffney et al., 1996; Perry et al., 2021) and fulvic acids (lower molecular weight, soluble at all pH values) vary amongst sites. Upstream waters have the most recently formed DOC with the highest humic acid content, and as the water flows downstream, the molecules become older, more degraded by bacterial action and photo-oxidation, and therefore smaller, and there is the addition of more autochthonous (see Glossary) molecules from phytoplankton and bacteria.

There is now considerable evidence (reviewed by Gonzalez et al., 2006; Wood et al., 2011; Nelson, 2015; Morris et al., 2021) that DOC has positive effects on ionoregulation in fish, especially at low pH. Indeed, the presence of the very molecules that cause the acidity of black waters (Kullberg et al., 1993) may be a key factor also protecting endemic fish against the negative effects of low pH (Gonzalez et al., 1998, 2002; Matsuo and Val, 2007; Wood et al., 2003). Herein lies an as yet unsolved experimental conundrum – how can we separate the effects of low pH from those of high DOC? The use of strong mineral acids (e.g. HNO₃ HCl, H₂SO₄) to experimentally lower pH is clearly not the same as adding natural DOC, and creates an unnatural water chemistry with elevated concentrations of strong anions, while P_CO2 levels needed to reach pH 4.0 would be both unnatural and likely fatal to most species.

Although protection by DOC against low pH pathology was first shown with Amazon DOC and species, it is now clear that the phenomenon extends to species and DOC from other parts of the world (Al-Reasi et al., 2016; Duarte et al., 2016, 2018; Galvez et al., 2008; Matsuo et al., 2004). DOC also protects against metal toxicity by binding and removing the free cationic forms of metal from solution, thereby decreasing their bioavailability (Niyogi and Wood, 2004; Al-Reasi et al., 2011; Morris et al., 2021). This is undoubtedly the case in Amazon waters, where substantial concentrations of naturally occurring Fe, Al, Mn and Hg are complexed by DOC (Aucour et al., 2003; Furch, 1984; Holland et al., 2017; Silva et al., 2009; Wood et al., 2003), as well as contaminant metals such as Cu (Rocha et al., 1999). Cu is an ionoregulatory toxicant and experimental additions of Amazonia DOC reduced Cu toxicity to fish (Barbosa et al., 2019; Crémazy et al., 2016; Duarte et al., 2009). Crémazy et al. (2016) demonstrated through geochemical modelling that only part of the toxicity reduction could be explained by Cu–DOC complexation and attributed the other portion to the supportive physiological effects of DOC on ionoregulatory homeostasis. Future environmental acidification may release metal cations from natural DOC, causing toxicity to native organisms (Tipping, 1998).

Morris et al. (2021) have reviewed the possible mechanisms by which natural DOCs may promote ionoregulatory homeostasis and protect against acid toxicity. These include: (i) a Ca²⁺-like action to ‘tighten’ gill membrane permeability (Wood et al., 2003); (ii) making the gill TEP more negative, thereby opposing the depolarizing effect of low pH and favouring net Na⁺ uptake (Galvez et al., 2008; Sadauskas-Henrique et al., 2019); (iii) maintaining the coupling of Na⁺ uptake to ammonia excretion (Duarte et al., 2016; Wood et al., 2003); (iv) buffering the external micro-environment of the gills so as to minimize the inhibitory effects of low pH on apical Na⁺ uptake transporters (Al-Reasi et al., 2016; Wood et al., 2003); and (v) increasing the J_max values of the branchial ion uptake systems (Al-Reasi et al., 2016; Glover et al., 2005; Glover and Wood, 2005; Matsuo et al., 2004; Matsuo and Val, 2007; Wood et al., 2003), perhaps owing to increased Na⁺,K⁺-ATPase activity in the gills (McGeer et al., 2002). These actions may be both acute and chronic, and DOC may bind directly to gill surfaces (Campbell et al., 1997). Indeed, after 7 days of pre-exposure to DOC, the protective effects against acute low pH persisted even when the DOC was removed from the water (Duarte et al., 2016, 2018).

The properties of DOC that exert these physiological actions appear to be the same as those protecting against metal toxicity (Wood et al., 2011). Al-Reasi et al. (2013) have related both phenomena to an optical property, the specific absorbance coefficient at 340 nm (SAC340, an indicator of aromaticity) and a chemical property, the proton binding index (PBI, an indicator of reactivity). These are greatest for large, optically dark humic acid molecules with abundant phenolic groups, of allochthonous origin. Therefore, as DOC levels increase with climate change, whether the effects are beneficial to fish will depend not only on concentration, but also on the chemical structure of the DOCs. The DOCs mobilized from deforested areas will probably be composed of older, smaller molecules that are less aromatic and more labile to microbial degradation than DOC exported from forested areas (Drake et al., 2019). Therefore, we predict that DOC from deforested areas will be less beneficial to fish. Furthermore, not all DOC is protective or benign; some, such as commercially available humic acid from peat or coal (Wood et al., 2003), as well as natural allochthonous DOC from the upper Rio Negro that has degraded during storage (Sadauskas-Henrique et al., 2019), may actually increase ionoregulatory disturbance at low pH.

Deforestation and agriculturalization will be accompanied by increased erosion, siltation and turbidity (Figueiredo et al., 2010; Leitão et al., 2018). The concentration of suspended particles in white waters is much higher (>10-fold) than in black or clear waters (Crémazy et al., 2019; Holland et al., 2017), so the species native to white waters will likely be more resistant. Because particles tend to settle, benthic species appear to be most affected, and there is evidence of associated shifts in species assemblages in silt-impacted areas of the Amazon, opening up opportunities for exotic invaders (Leitão et al., 2018). In temperate species, a variety of lethal and sublethal effects of increased particles have been reported, including behavioural (e.g. avoidance, disruption of predator–prey activities and mating), respiratory (e.g. suffocation, increased O₂ consumption, decreased swimming performance), reproductive (impacts on spawning and early life stages) and generalized stress responses (Kjelland et al., 2015). Ionoregulatory impacts have been reported in black-water species exposed to particle-rich white water, perhaps associated with physical damage to the gills (Crémazy et al., 2019). However, on the positive side, particles can complex cationic free Cu²⁺ out of solution with great efficiency, thereby reducing toxicity (Crémazy et al., 2019; Dal Pont et al., 2017). The same is probably true for other metals (Seyler and Boaventura, 2003).

In the next century, water temperatures (currently 28–32°C) in Amazonia are predicted to increase by 2.2 to 7.0°C (IPCC, 2014) and this may be exacerbated by shade removal by deforestation (Ilha et al., 2018; Leitão et al., 2018). Unless there is rapid evolutionary change, the consequences for fish appear dire. A recent study (Jung et al., 2020) on 37 Amazonian species found that the acute upper lethal temperatures (CT_max) ranged from 36.1 to 42.8°C. Acclimation to higher temperature caused modest increases in CT_max but decreases in hypoxia tolerance. In a subset of 13 species, the chronic (4 weeks) upper lethal temperatures (CL_max) were typically 6–9°C below the CT_max values. This means that long-term increases in temperature only slightly above current river temperatures (28–31°C) may eliminate a large number of Amazonian fish species. Overall, there was a positive interspecific correlation between CT_max and acute hypoxia tolerance, and on an intraspecific basis, small acute increases in temperature resulted in decreases in hypoxia tolerance, whereas acute hyperoxia exposure caused small increases in CT_max (Jung et al., 2020). Furthermore, low factorial aerobic scope (associated with high routine metabolic rate) was correlated with low CT_max on an interspecific basis (Campos et al., 2017). In a comparison of two closely related Characiform species, greater thermal tolerance was associated with lesser temperature sensitivity of routine metabolic rate, lower critical oxygen tension (Campos et al., 2017) and greater capacity for anaerobic metabolism (Fé-Gonçalves et al., 2018). From these and other studies (e.g. LaPointe et al., 2018), it is clear that there is great inter-species variability. Nevertheless, in total, these observations suggest an interaction between thermal tolerance and the O₂ delivery cascade, such that climate warming will favour species with low routine metabolic rates, low temperature sensitivity of routine metabolic rates, high anaerobic capacity and high hypoxia tolerance.

Thermal effects on growth rate in the laboratory are variable (e.g. Oliveira and Val, 2017; LaPointe et al., 2018). However, in the field, there is a general trend for smaller mean body size of fish at the population level in streams that have recently warmed owing to deforestation (Ilha et al., 2018), a manifestation of Bergmann's rule (Bergmann, 1847) over very short temporal and spatial scales. In the Amazon context, the causative factors for smaller body size are likely multifactorial, as with many of the other effects of increased temperature (Ilha et al., 2018).

Indeed, as with hypoxia, the effects of increased temperature will likely be additive or synergistic with a variety of other stressors – for example, the effect of both low pH and elevated temperature in decreasing blood oxygen affinity (Val et al., 2016). The pathological effects of combined, realistic elevations in both atmospheric P_CO2 and temperature, equilibrated with the water holding the fish in special climatic chambers, have now been demonstrated for many parameters. These include Cu-induced oxidative stress (Braz-Mota et al., 2017), the intensity of parasitic infections (Costa and Val, 2020) and the prevalence of skeletal deformities (Lopes et al., 2018).

Most physiologists would consider a water P_CO2 of approximately 0.04 kPa (0.3 Torr) to represent the normal situation. However, for Amazon fishes, well-aerated laboratory ‘control’ conditions may be unrealistic as their natural waters are strongly supersaturated with CO₂, owing to the respiratory breakdown of organic materials. Reported values include a yearly mainstream average of 0.44 kPa (3.3 Torr) in major rivers, 0.51–2.03 kPa (3.8–15.2 Torr) in forest streams, 0.17–0.48 kPa (1.3–3.6 Torr) in the black-water Rio Negro, 0.35–0.72 kPa (2.6–5.4 Torr) in the white-water Rio Solimões, 0.03 –0.36 kPa (0.2–2.7 Torr) in several clear-water rivers and >8 kPa (>60 Torr) in floodplains (Furch and Junk, 1997; Rasera et al., 2013; Richey et al., 2002). The worst-case climate change scenarios 0.13 kPa (≤1 Torr) elevation in atmospheric P_CO2), while significant, will be superimposed on natural aquatic P_CO2 levels that are already much higher.

Therefore, many Amazon fish are pre-adapted for high P_CO2 tolerance. Equilibration of environmental CO₂ across the gills is unavoidable, so blood levels are always above water levels. However, air-breathers such as the South American lungfish may be able to counteract CO₂ entry from the water by excretion through the air-breathing organ, thereby controlling the rise in blood P_CO2 levels (which are already high because of air-breathing; Dejours, 1975), so as to keep them below waterborne levels (Sanchez et al., 2005).

Many Amazon fish prioritize the regulation of intracellular over extracellular pH (Brauner et al., 2004b; Harter et al., 2014; Heisler, 1982). When experimentally challenged with high waterborne P_CO2, blood pH drops immediately, with little subsequent compensation by extracellular HCO₃⁻ accumulation, whereas tissue pH is regulated rapidly and precisely. This is very different from the classic ‘coupled regulation’ seen in model temperate species, where extracellular and intracellular pH are regulated in tandem (Brauner and Baker, 2009). Possible explanations for this strategy include: (i) the dilute, acidic, weakly buffered Amazon waters are a poor source of HCO₃⁻, and a difficult medium into which to excrete H⁺, compounded by low Na⁺ and Cl⁻ as counterions for branchial acid–base transporters (Larsen and Jensen, 1997); (ii) the intracellular compartment is much better buffered than the extracellular compartment, so intracellular pH correction requires far less HCO₃⁻; (iii) this small amount of HCO₃⁻ can be moved easily from plasma to tissues; and (iv) the HCO₃⁻ threshold argument that it is impossible to raise plasma [HCO₃⁻] above approximately 30 mmol l⁻¹, because it will drive plasma [Cl⁻] too low (Heisler, 1986). The absolute limit is debatable, but it certainly would prevent complete compensation at water P_CO2 ≥2.67 kPa (≥20 Torr), which occurs frequently in Amazonia. Shartau et al. (2016) have argued that preferential regulation of intracellular pH is an exaptation for air-breathing, which is accompanied by large elevations in blood P_CO2 (Dejours, 1975). Indeed, the Amazonian species exhibiting preferential intracellular pH regulation all have some air-breathing capacity (Brauner et al., 2004b; Harter et al., 2014; Heisler, 1982; Shartau et al., 2016).

Intermittent aquatic hypoxia is one of the most important evolutionary and ecological drivers in Amazonia and will become more intense with global change. Most Amazon fish are already adapted to some degree to cope with hypoxia at the levels of the tissues (e.g. high glycolytic capacity and/or ability to turn down metabolic rate; Muusze et al., 1998; Almeida-Val et al., 2000) and blood O₂ transport (e.g. blood–O₂ affinities which can be quickly increased; Val et al., 2015; Scott et al., 2017), and many have a capacity for aquatic surface respiration and/or air-breathing (Graham, 1997; Randall et al., 1981; Val and Almeida-Val, 1995). Organs that have been adapted for these purposes include the swim bladder, gills, lips, buccal cavity, skin, fins and gastrointestinal tract. In most Amazon fish, regulation of haemoglobin–O₂ affinity is accomplished by a routinely higher concentration of GTP, which is a stronger modulator than ATP, commonly found in temperate zone fishes. However, several other organic phosphates (2,3DPG, IP2, IPP) are also found, and are regulated according to environmental challenges (Marcon et al., 1999; Val, 2000; Weber et al., 2000). At first glance, those species that have the greatest capacity for O₂ uptake from air, such as the obligate air-breathing pirarucu (Pelster et al., 2020a), would seem best-equipped to cope. However, the early life stages are obligate water-breathers (Brauner et al., 2004a). Furthermore, throughout life they must continue to breathe water for CO₂ and ammonia excretion, ionoregulation and acid–base regulation (Frommel et al., 2021; Gonzalez et al., 2010). Therefore, gill function will continue to be impacted by aquatic hypoxia. Indeed, the pirarucu (Wood et al., 2020), like several other hypoxia-tolerant Amazonian species (Robertson et al., 2015), displays the exact opposite of the standard osmorespiratory compromise (Randall et al., 1972; Wood and Eom, 2021). Gill ion fluxes quickly decrease rather than increase during hypoxia. In the oscar (De Boeck et al., 2013; Wood et al., 2007, 2009), this is achieved by rapid extension of pavement cells over gill ionocytes (Matey et al., 2011), thereby reducing ion and water permeability without impeding O₂ transfer (Scott et al., 2008). The short-term benefits for osmoregulation are obvious, but longer-term impacts (impeded ammonia and acid–base regulation) may be costly. Additional costs associated with increased reliance on air-breathing include the potential for aerially sourced O₂ to be lost at the gills during aquatic hypoxia (Scott et al., 2017; Aaskov et al., 2022), greater oxidative stress in the air-breathing tissues (Pelster et al., 2016, 2018, 2020b; Pelster and Wood, 2018), increased risk from avian predation when surfacing (Sloman et al., 2009, 2006) and increased reliance on the diet for ion acquisition (Pelster and Wood, 2018; Wood et al., 2016).

Given these costs, it may well be that some predominantly water-breathing species will be favoured in a future more hypoxic Amazon, especially those with high glycolytic capacity and/or great ability to reduce metabolic rate during hypoxia, such as the oscar (Almeida-Val et al., 2000; Chippari-Gomes et al., 2005; Muusze et al., 1998). Additional specific survival mechanisms may include skimming with a protruded lip as seen in matrincha and tambaqui (Anjos et al., 2008; Rantin and Kalinin, 1996; Val and Almeida-Val, 1995; Val et al., 1998), an ability to oxygenate the nests as in angelfish (Korzelecka-Orkisz et al., 2012), the ability to rapidly find normoxic refugia as in the lateral migrations (Cox-Fernandes, 1997; Silva and Stewart, 2017; Val and Almeida-Val, 1995) and the school migrations near the surface that help oxygenate the water (A.L.V., personal observation).

The fish of Amazonia evolved under continuous environmental change. Throughout the evolutionary process, climate change has acted additively, causing the emergence of adaptations, speciation and extinctions, already well documented among fish (Graham, 1997; Somero, 2010; Val et al., 1998). Therefore, many native species can live at the extreme limits of environmental challenges, particularly with respect to water temperature, low pH, hypoxia and even anoxic conditions. Climate change is now pushing many fish species to almost physiologically unsustainable conditions. This situation has different contours in different ecosystems, such as in the estuaries of Australia (Scanes et al., 2020). Furthermore, these synergistic interactions of climate change with other environmental features have specific effects on local fauna (Pörtner et al., 2021), including ichthyofauna. As already mentioned, the responses of different fish species and fish from different regions to environmental variations are not homogeneous (there is a vast literature on this; see Brauner et al., 2019; Farrell and Richards, 2009; Lefreve et al., 2021). Besides direct effects such as temperature and P_CO2 increase, associated specific factors also affect Amazonian fishes: increased amplitudes of flood pulses, and changes in vegetation cover, siltation and DOC characteristics (Fig. 4).

At least five classes of climate change effects on fish can be recognized (Table 2). These may occur simultaneously and amplify the effects of a particular parameter. Note that synergistic effects of climate change on fish can be either additive or interactive. Controlled laboratory simulations of future climate conditions indicate that synergistic effects will impose new challenges to Amazon fish, different from the effects of a given single environmental condition. Some species such as tambaqui seem to be resilient (Val and Oliveira, 2021), while others are very sensitive. It is worth noting that in their natural environments, although physiological effects can be attenuated or amplified, animals have a range of possibilities to defend themselves biologically, which is not the case under experimental conditions. Thus, long-term studies in the natural environment involving life history, predation and other ecological factors are needed to predict the future resilience of different species.

The major fish groups of Amazonia appeared during geological periods that had very high CO₂ and low O₂ relative to current levels. Will present and future populations be able to use information hidden in their genome that allowed them to survive these conditions in the past? Conceivably this evolutionary history could favour their persistence in the face of current and future climate change scenarios. Our Review suggests that climate change will favour predominantly water-breathing species with low routine metabolic rates, low temperature sensitivity of routine metabolic rates, high anaerobic capacity, and high thermal and hypoxia tolerance. Water temperature may be a particularly decisive factor. Amazon fish appear to have reached an evolutionary point of very high sensitivity to temperature, as they now live very close to their critical thermal limits. This may be a turning point for the whole fish diversity of Amazonia, resulting in many local extinctions and yet through intense selection pressures, the appearance of new adaptations to extreme synergistic challenges. For now, so few species have been studied out of the immense diversity of fishes in the region that any generalizations will probably overlook other important biological parameters. Over the next few decades, the many unanswered questions will call for integrated analysis, and physiologists and experimental biologists, among others, will be able to study evolutionary change in real time in Amazonia.

Many people helped us in the field work in the Amazon and in the laboratory. It is on behalf of two of them, Maria de Nazaré Paula da Silva and the late fisherman and field technician Agenor Negrão, that we extend our deepest thanks. Everyone's help made our work possible.