Ocean acidification does not limit squid metabolism via blood oxygen supply

Atmospheric carbon dioxide (CO₂) partial pressure (P_CO2) has increased from the pre-industrial mean of 28 Pa (280 μatm, ppmv) to over 40 Pa (≈400 μatm) today (Caldeira and Wickett, 2005) and may reach 100 Pa (1000 μatm) by the year 2100 (IPCC, 2014). Elevated environmental P_CO2 will influence marine organisms indirectly via global warming. However, anthropogenic CO₂ also diffuses into the ocean, where it reacts with water, resulting in reduced pH. This phenomenon, known as ocean acidification (OA), may affect animal performance in numerous ways (Fabry et al., 2008; Clements and Hunt, 2015). For example, it has been proposed that OA may impair the oxygen (O₂) supply capacity of marine animals via its effect on pH-sensitive respiratory proteins (Widdicombe and Spicer, 2008; Fabry et al., 2008; Pörtner, 2012; Miller et al., 2016; Seibel, 2016). Even small losses in O₂ supply capacity could hinder an animal's exercise performance or environmental hypoxia tolerance.

Shallow-water active squid, such as those in the families Loliginidae and Ommastrephidae, are ideal study organisms for examining impacts of OA on O₂ supply because their respiratory proteins (hemocyanins) are among the most pH sensitive of any marine animal (Brix et al., 1989; Bridges, 1994; Pörtner and Reipschläger, 1996; Seibel, 2016). Sensitivity to pH, quantified as the Bohr coefficient (Bohr et al., 1904), is optimal for O₂ delivery to the tissues at half the respiratory quotient (Lapennas, 1983), which would be between −0.35 and −0.5 in cephalopods. Squid hemocyanin, however, often has a Bohr coefficient of <−1 (Bridges, 1994). The extreme sensitivity in cephalopods may result in large impairments in blood–O₂ binding affinity from relatively small changes in blood pH.

Furthermore, hemocyanins are not contained within red blood cells, but are freely dissolved in the blood, limiting their concentration owing to viscosity and osmotic constraints. Unlike fishes or invertebrates with red blood cells, which can increase hemoglobin concentration or hematocrit to increase O₂ supply (Johansen and Weber, 1976), squid are constrained in their blood O₂ carrying capacity, and are fully dependent on their cardiovascular system for O₂ delivery (Birk et al., 2018). Squid are thought to utilize most of the O₂ available in their blood with very little venous reserve even under resting conditions (Wells, 1992; Pörtner, 1994). Cephalopods, unlike fishes and crustaceans, are not known to rely on organic cofactors such as adenosine phosphates and lactate to modify hemocyanin–O₂ affinity (Johansen and Weber, 1976; Mangum, 1997).

Such physiological considerations have led to the concern that, in the absence of acclimation or adaptation, squid metabolism may be strongly affected by OA (Pecl and Jackson, 2007; Seibel, 2016). In fact, Pörtner (1990) estimated that a 0.1–0.15 unit decrease in arterial pH would be lethal for active squid. Redfield and Goodkind (1929) found that blood–O₂ transport in the loliginid squid Doryteuthis pealeii was impaired by acute exposures (10–15 min) to P_CO2 levels up to 3200 Pa (31,500 μatm CO₂), resulting in death. Rosa and Seibel (2008) reported reduced metabolic rate and activity at much more modest CO₂ levels (100 Pa; 1000 μatm), which they attributed to the high pH sensitivity of hemocyanin in Dosidicus gigas. Similar results have been found in embryonic squid exposed to 170 Pa (1650 μatm) CO₂ (Rosa et al., 2014).

However, Hu et al. (2014) found no effect of 160 Pa (1600 μatm) CO₂ on metabolism even after 1 week of exposure in the loliginid squid Sepioteuthis lessoniana. Cuttlefish, which have lower O₂ demand but hemocyanin with similarly high pH sensitivity, also exhibited no effect on metabolism after 24 h or growth rate over 40 days at P_CO2 levels up to 615 Pa (6100 μatm) (Gutowska et al., 2008). Such tolerances may be attributed to the high capacity for blood acid–base regulation in most cephalopods (Melzner et al., 2009; Hu and Tseng, 2017). The studies to date are not directly comparable, each having employed a different species, P_CO2 level, exposure duration and method. Thus, the variable results are perhaps not surprising. A mechanistic physiology-based model can help elucidate these diverse whole-animal metabolic responses to CO₂ reported in the literature.

Although all loliginid and ommastrephid squid have rather active lifestyles, individual species have evolved in very different environments that may select for quite different CO₂ or hypoxia tolerances. For example, Dosidicus gigas is an ommastrephid squid that inhabits the eastern tropical Pacific, where a pronounced oxygen minimum zone (OMZ) exists. They encounter strong gradients in P_O2 (Δ>10 kPa), P_CO2 (Δ≥100 Pa; 1000 μatm) and temperature (Δ>10°C) during their daily migration into the OMZ (Gilly et al., 2006, 2012; Franco et al., 2014). The squid suppress total metabolism by 50% while in the core of the OMZ during daytime hours (Seibel et al., 2014). In contrast, the loliginid squid Doryteuthis pealeii inhabits coastal and shelf waters in the western Atlantic, and never encounters such extreme hypoxia or hypercapnia [though bays can reach P_CO2>50 Pa (500 μatm) in the summer months; Turner, 2015]. As such, D. gigas is adapted to more extreme environmental conditions than D. pealeii.

In this study, we examined the effects of CO₂ on hypoxia tolerance in two squid species with similar O₂ demands but differing hypoxia tolerances, D. gigas and D. pealeii, to determine what impact OA may have on O₂ supply in squid. We applied two independent approaches to this question. First, we conducted laboratory experiments to examine the effect of CO₂ on hypoxia tolerance. Second, we constructed a model of blood acid–base balance and O₂ delivery with variable inputs for blood–O₂ affinity, CO₂ sensitivity of hemocyanin and buffering capacity to predict the physiological changes in O₂ supply expected by end-of-the-century OA.

Adult and juvenile Dosidicus gigas (D'Orbigny, 1835; n=16) were jigged at night in Guaymas Basin, Gulf of California, Mexico, from 16 May 2015 to 1 June 2015 aboard the R/V Oceanus. Juvenile and adult Doryteuthis pealeii (Lesueur, 1821; n=29) were caught in southern Narragansett Bay, RI, USA, by either hand jigs or, less commonly, benthic otter trawl in April through November 2014–2016. Morphometrics of both species are shown in Table 1. Dosidicus gigas were placed immediately in a respirometer aboard ship for acclimation, whereas D. pealeii were transported in an aerated cooler to the Durbin Aquarium facility at the University of Rhode Island, where they were held in tanks of at least 540 liters with flow-through filtered seawater. Doryteuthis pealeii were fed grass shrimp (Palaemonetes sp.) or 1 cm wide herring (Clupea harengus) steaks ad libitum during the holding period before experiments were conducted. Prior to acclimation and experimental trials with D. pealeii, temperature was maintained at 15°C (within 5°C of capture temperature) and P_CO2 varied with ambient conditions in Narragansett Bay, where P_CO2 typically ranges from 10 to 70 Pa (100–700 μatm; Turner, 2015).

Hypoxia tolerance was assessed by measuring O₂ consumption rates of individual squid under progressively declining seawater P_O2 using intermittent respirometry. All experiments were conducted in a 90 liter swim tunnel respirometer (Loligo Systems, Viborg, Denmark) with a 70×20×20 cm working section in which the animal was confined. Trials were conducted at surface pressure, as hydrostatic pressure has little effect on metabolism in squid (Belman, 1978). Acclimation and trials were conducted at 15°C for D. pealeii. However, for D. gigas, temperature was maintained at ambient sea surface temperature, which varied from 23 to 27°C. Measurements were adjusted accordingly (see below). CO₂ treatment began immediately upon placing the animal in the respirometer. Median acclimation duration was 10 and 13 h for D. gigas and D. pealeii, respectively, but varied from 8 to 13 h for D. gigas and from 5 to 59 h for D. pealeii. This allowed time for thermal acclimation (for D. pealeii), completion of digestion from any previous meal (Wells et al., 1983; Katsanevakis et al., 2005), acclimation to CO₂ conditions (Gutowska et al., 2010a), and recovery from handling stress. Oxygen consumption rates from squid in swim-tunnel experiments extrapolated to zero activity are similar to rates measured after 6 h acclimation (Seibel, 2007). Previous measurements (Trueblood and Seibel, 2013) have shown that D. gigas stabilize metabolic rate by 6 h after enclosure in respirometers. Animals were free to move within the working section of the respirometer, which was 3–4 mantle lengths long. Thus, we refer to the metabolic rates measured here as routine metabolic rates (RMR) rather than standard metabolic rates (SMR), although the animals often rested on the bottom of the chamber.

After acclimation for each animal, the respirometer was closed and the P_O2 was drawn down by the animal. Every 4–5 kPa O₂ (every 1–7 h depending on the rate of metabolism), the respirometer was flushed with seawater at matching P_O2 and experimental P_CO2 to minimize NH₃ and CO₂ accumulation. The average flush provided a 70% water exchange, based on estimated mixing efficiency from Steffensen (1989). P_O2 was measured every 10 s with an O₂-sensitive spot (Fibox 3 meter and PSt3 spots; PreSens Precision Sensing GmbH, Regensburg, Germany). The oxygen meter was calibrated with air-saturated seawater and concentrated NaSO₃ solution (P_O2=0). Water velocity inside the respirometer was kept low (≈5 cm s⁻¹) to allow homogeneous mixing.

Flush water was UV-treated, then stored in a reservoir, brought to treatment temperature and bubbled with pure nitrogen gas to draw down dissolved P_O2. To produce high CO₂ conditions, pure CO₂ gas (AirGas ‘Bone Dry’ grade, Radnor, PA, USA) was dispersed through a peristaltic pump and bubbled into the intake of a submersible aquarium pump to enhance dissolution (Jokiel et al., 2014). The treated reservoir water was then flushed through the respirometer. The appropriate volume of CO₂ gas added for each flush was calculated using the R package ‘respirometry’ (https://cran.r-project.org/web/packages/respirometry/index.html). Water samples were collected from the respirometer output at the start of each flush for carbonate chemistry analyses. For the D. pealeii trials, water samples were also analyzed from the incoming flush water.

Metabolic rate (Ṁ_O2) was monitored in real time and the trial was ended soon after the animal reached P_crit, the environmental P_O2 below which aerobic metabolism (indicated by O₂ consumption rate) decreases. The average duration of the trials after acclimation was 7 and 23 h for D. gigas and D. pealeii, respectively, owing to temperature and animal size. This resulted in an average total exposure duration to treatment conditions by the time the animal reached P_crit of 17 and 36 h, respectively. At the end of each trial, the animal was removed and the ‘background’ O₂ consumption rate of the microbial community in the respirometer was measured and deducted from calculated squid Ṁ_O2. Gill length relative to dorsal mantle length (DML) was measured in both species.

During 17 of the D. pealeii trials, the animals were filmed for 1 min every 30 min to monitor ventilation rate. The camera was placed above the respirometer and a mirror was placed at a 45 deg angle to the camera, allowing simultaneous monitoring of the animal from a dorsal and lateral view.

As hypoxia progressed, three possible effects on ventilation rate were considered: (1) ventilation rate is unaffected by hypoxia, (2) ventilation rate increases linearly with progressive hypoxia and (3) ventilation rate is unaffected at moderate P_O2 levels but increases at more extreme hypoxia (i.e. breakpoint relationship). A model was fit for each of these relationships using maximum likelihood estimation with normal error distributions with the mle2() function from the R package ‘bbmle’ (https://cran.r-project.org/web/packages/bbmle/index.html). The best-fitting model was chosen using the Bayesian information criterion (BIC). Effects of P_CO2 on ventilation rate were assessed by linear regression based on the estimated CO₂ level of each ventilatory observation.

Ṁ_O2 was calculated from the slope of a linear regression of P_O2 over time. The number and quality of Ṁ_O2 measurements obtained from this technique are dependent on the width of the time bins used. The time bin width scaled with P_O2 such that the time bins at high O₂ covered 1/10th the trial duration and the time bins at low O₂ covered 1/100th the trial duration. This provided an optimal balance between precision and resolution throughout each trial. Ṁ_O2 measurements derived from regressions with an R²<0.7 were discarded.

To calculate P_crit, a traditional breakpoint relationship was fit using the segmented() function from the ‘segmented’ R package (Muggeo, 2008), which fits a broken-stick regression to the relationship between Ṁ_O2 and P_O2. Then, a 95% prediction interval was added around the oxyregulating line to encapsulate a space in which all observed Ṁ_O2 values can reasonably be considered within the oxyregulating space. The ‘sub-PI’ P_crit is defined as the P_O2 at which the oxyconforming line intersects the lower limit of the 95% prediction interval (see Figs S1 and S2). This sub-PI method resulted in a lower variability in P_crit measurements than the traditional breakpoint method.

Trials in which the animal did not exhibit a clear breakpoint response to P_O2 were removed from P_crit analyses. Only Ṁ_O2 measurements with mean P_O2>P_crit were considered when calculating mean RMR. Median RMR and the lowest 10% of Ṁ_O2 observations for each individual were also estimated and gave similar results. For trials where no P_crit could be reliably established, RMR was determined as the mean of the Ṁ_O2 values.

where k₁ and k₂ are the calculated values (e.g. Ṁ_O2 or P_crit) measured at temperatures T₁ and T₂, respectively. Typical Q₁₀ values for metabolic rate in ectotherms range from 2 to 3 (Hochachka and Somero, 2002), meaning that metabolic rate doubles or triples with a 10°C increase in temperature. The Q₁₀ was calculated using the Q10() function from the R package ‘respirometry’, and bootstrap bias-corrected and accelerated confidence intervals were fit to form confidence bands.

Seawater carbonate chemistry was assessed by measuring pH (total scale) and total alkalinity (TA) from water entering and expelled from the respirometer during acclimation and flushes. pH was measured spectrophotometrically at 25°C with m-Cresol Purple, a pH-sensitive dye (Clayton and Byrne, 1993), using standard operating procedure (SOP) 6b from Dickson et al. (2007), modified for use with a 1 cm path length cuvette. Based on pH measurements from flush water samples, seawater pH inside the respirometer during the inter-flush periods of the trials was calculated using the predict_pH() function from the R package ‘respirometry’. A respiratory quotient (RQ; ratio of CO₂ produced to O₂ consumed) of 0.85 was used because cephalopods mainly utilize protein catabolism (Hoeger et al., 1987). For the D. pealeii trials, where water samples were analyzed from both respirometer input and output at every flush, pH was calculated from both the start and end of each inter-flush measurement period. The values from these two methods of calculation differed by only 0.06 pH units on average, which corroborates this RQ value, and were averaged.

Alkalinity was measured either by SOP 3b from Dickson et al. (2007; potentiometric titration) or by Liu et al. (2015; spectrophotometric titration). Alkalinity measurements were calibrated with certified reference materials provided by Andrew Dickson (Scripps Institution of Oceanography, La Jolla, CA, USA). P_CO2 was calculated from inter-flush pH and trial-averaged TA using carbonate dissociation constants from Lueker et al. (2000) via the R package ‘seacarb’ (https://cran.r-project.org/web/packages/seacarb/index.html). Mean P_CO2 values for each trial were chosen as the environmental metric for analysis. The P_CO2 at P_crit was also estimated and gave similar results. Because of the unavoidable CO₂ build-up when the respirometer was closed, seawater pH during each trial varied within the respirometer over a median range of 0.21 and 0.28 pH units in the D. gigas and D. pealeii trials, respectively.

A physiological model was developed to estimate the magnitude of an effect seawater P_CO2 has on blood O₂ transport and P_crit in squid. This model intentionally does not include any ability for blood acid–base compensation, a well-developed trait in squid that makes it highly unlikely that OA would result in long-term blood acidosis (Melzner et al., 2009; Hu and Tseng, 2017), for two reasons: (1) we have no specific knowledge on the rate or extent of acid–base compensation in either species studied, and (2) by removing the ability to compensate for blood acidosis in this model, a reasonable upper bound on expected hypercapnic impact can be assessed. Physiological parameters [normocapnic blood P_CO2 and pH, non-bicarbonate buffering capacity (β_NB), Bohr and Hill coefficients, hemocyanin P₅₀, arterial P_O2 at normoxia, and P_crit] were collected from the literature for D. pealeii and D. gigas (Table 2).

This has been observed in cephalopods (Gutowska et al., 2010a; Hu et al., 2014; Häfker, 2012) as well as fishes exposed to hypercapnia (Janssen and Randall, 1975; Strobel et al., 2012a; Esbaugh et al., 2012, 2016; Ern and Esbaugh, 2016; for a review, see Heuer and Grosell, 2014). In the absence of any change in metabolic rate or CO₂ production, the flux rate must remain constant if gradual respiratory acidosis is to be avoided. Thus, an increase in seawater P_CO2 elevates blood P_CO2 and reduces blood pH. CO₂ solubility and dissociation constants for seawater (S=35) from Lueker et al. (2000) were used. These constants are similar to values calculated in crab hemolymph (Truchot, 1976), which has ionic properties similar to those of squid blood.

According to Eqn 4, as blood P₅₀ increases due to acidosis, the arterial P_O2 necessary to maintain the same Hc–O₂ saturation increases as well in a nonlinear relationship. As long as O₂ demand is unchanged (e.g. no change in temperature or physiological activity), P_crit is reached when Hc–O₂ saturation falls below a set threshold (Redfield and Goodkind, 1929; Speers-Roesch et al., 2012) at which point the amount of O₂ carried in the blood is insufficient to support cellular metabolism.

The change in arterial P_O2 for a given change in environmental P_O2 can be calculated from the arterial P_O2 under environmental air saturation and under anoxia assuming a linear relationship between environmental and arterial P_O2 (Eddy, 1974; Johansen et al., 1982; Houlihan et al., 1982; Speers-Roesch et al., 2012). The increase in arterial P_O2 required to reach the Hc–O₂ saturation threshold under hypercapnia translates to an increase in P_crit. Blood acid–base compensation was intentionally left out of the model because acid–base regulatory ability can minimize sensitivity to OA (Melzner et al., 2009). In so doing, a reasonable upper bound on expected hypercapnic impact can be assessed.

Models were run for both species at a range of temperatures using physiological parameters from the literature from these or closely related species (Table 2). Blood pH was temperature-adjusted by −0.02°C⁻¹ to match in vivo temperature dependence of blood pH (Howell and Gilbert, 1976). The Bohr coefficient, P₅₀ and P_crit were temperature-adjusted using species-specific temperature relationships (Table 2). The temperature dependence of P_crit in D. pealeii has not been measured to date. Therefore, the Q₁₀ value for hemocyanin P₅₀ was applied to P_crit in D. pealeii because of the close correlation between these parameters (Mandic et al., 2009).

To assess the model's reliability, the model was run with published physiological parameters for D. pealeii (Table 2) and compared with empirical data from Redfield and Goodkind (1929). They examined the effects of acute (10–15 min) seawater hypercapnia (up to 3200 Pa or 31,500 μatm CO­₂) on lethal P_O2 in D. pealeii. Blood acid–base compensation was not incorporated to the model due to the acute CO₂ exposures (10–15 min) by Redfield and Goodkind (1929).

The model was also run for both species at temperatures ranging from 0 to 25°C. Three levels of environmental hypercapnia were also considered. First was a rise in P_CO2 of 60 Pa (≈600 μatm) as expected for the mean sea surface. Given a ΔP_CO2 of 60 Pa for air-equilibrated seawater, CO₂ dissociation constants from Lueker et al. (2000), average ocean TA (Lee et al., 2006) and an RQ of 0.75, environmental P_CO2 may be expected to rise by ≈130 Pa (≈1300 μatm) in regions with 50% air saturation, and ≈200 Pa (≈2000 μatm) in regions with 10% air saturation. These ΔP_CO2 levels were examined to cover a broad range of conditions that loliginid and ommastrephid squid may encounter in future oceans (Melzner et al., 2013).

Experimental seawater parameters from the trials are shown in Table 3. All results are expressed as means±s.d.

There was no effect of hypercapnia on metabolic rate for either species (D. pealeii: t₁₇=−1.08, P=0.297; temperature-adjusted D. gigas: t₁₄=0.11, P=0.914; Fig. 1A). Dosidicus gigas had an RMR of 13.2±2.6 μmol O₂ g⁻¹ h⁻¹ at 22.5–26.6°C. Its rate increased significantly with temperature (t₁₄=1.6, P=0.123), with a Q₁₀ of 1.9 (Fig. 2A). Doryteuthis pealeii had an RMR of 6.5±2.5 μmol O₂ g⁻¹ h⁻¹ at 15°C. Metabolic rate was 1.8 μmol O₂ g⁻¹ h⁻¹ higher in Dosidicus gigas than in D. pealeii (t₃₅=3.31, P=0.002) once adjusted to a common temperature (15°C) using a Q₁₀ of 1.8 (derived from D. gigas Ṁ_O2 measurements from the literature spanning from 6.5 to 25°C; Gilly et al., 2006; Rosa and Seibel, 2008; Trübenbach et al., 2013; Trueblood and Seibel, 2013; Seibel et al., 2014; present study; Fig. 2C).

Seawater P_CO2 had no detectable effect on temperature-adjusted P_crit in either species (Fig. 1B). The P_crit for D. pealeii at 15°C was 3.9±0.8 kPa. Dosidicus gigas mean P_crit was 3.8±1.2 kPa, but it increased with temperature from 23°C to 27°C with a Q₁₀ of 1.8 (Fig. 2B). When adjusted to a common temperature (15°C) using a Q₁₀ of 1.8 (derived from D. gigas P_crit values from the literature ranging from 6.5 to 27°C; Gilly et al., 2006; Trueblood and Seibel, 2013; Fig. 2D), mean D. gigas P_crit was 1.4 kPa lower than D. pealeii P_crit (t₂₄=3.84, P<0.001).

Ventilation rate in D. pealeii had a breakpoint relationship with P_O2, remaining stable at high P_O2 (0.77 Hz) but increasing by 0.04 Hz kPa⁻¹ O₂ with progressive hypoxia below 9 kPa (Fig. 3A). Although O₂ strongly influenced ventilation, no effect of seawater P_CO2 was found on normoxic (P_O2>9 kPa) ventilation rate (t₁₄=−0.38, P=0.71; Fig. 3B). Dosidicus gigas relative gill length (38% of DML) was longer than that of D. pealeii (29% of DML; t₂₅=5.39, P<0.001).

Running the O₂ supply capacity model with physiological parameters from the literature for D. pealeii (Table 2) matched the empirical measurements of lethal P_O2 from Redfield and Goodkind (1929) very well even up to 3200 Pa or 31,500 μatm CO₂ (Fig. 4).

Both species experienced similar levels of arterial blood acidosis (Fig. 5A), with blood pH declining by 0.169 units at most under ΔP_CO2=200 Pa. Under more modest hypercapnia (ΔP_CO2=60 Pa), blood pH declined by 0.045 units, on average. Because of its lower hemocyanin–O₂ binding affinity (higher P₅₀), the O₂ binding curve for D. pealeii was more strongly impacted by blood acidosis than the curve for D. gigas, leading to a larger rise in P_crit and fall in arterial hemocyanin saturation from the same CO₂ exposure (Fig. 5B–D). For both species, the expected ΔP_crit was largely insensitive to temperature. The expected rise in P_crit from a 60 Pa increase in CO₂ was 0.52 and 0.24 kPa for D. pealeii and D. gigas, respectively. The arterial hemocyanin saturation was minimally affected in both species even at the highest hypercapnia exposure owing to the high arterial P_O2 relative to P₅₀.

We found no effect of seawater P_CO2 up to 143 Pa (1410 μatm) on metabolic rate or hypoxia tolerance in either species (Fig. 1). Rosa and Seibel (2008) had previously found that a P_CO2 of ≈100 Pa (1000 μatm) had caused a 20% decrease in D. gigas RMR relative to 30 Pa (300 μatm) and had attributed this suppression to hypercapnia-induced limitation to blood–O₂ binding. A 20% decline in RMR (black diamond in Fig. 1) fell outside of the 95% confidence band of the CO₂ effect on RMR observed here. There were a number of differences between the studies that may have produced the disparate effects, such as animal size, acclimation duration and measurement technique. It is possible that the previous study documented a short-term response to high CO₂ that is unrelated to blood O₂ supply. Rosa and Seibel (2008) measured metabolic rate for up to 6 h during exposure to CO₂, whereas the present study allowed the squid to acclimate to treatment CO₂ for at least 5 h and on average 10 h before beginning metabolic rate measurements.

Given what is now known about blood–O₂ binding in D. gigas (Seibel, 2013), impairment of O₂ supply could not have been the cause of the decline in metabolic rate and activity observed by Rosa and Seibel (2008). Based on the properties of D. gigas blood determined by Seibel (2013), the squid in the study by Rosa and Seibel (2008) should have had nearly completely O₂-saturated blood at all CO₂ levels encountered. Furthermore, if O₂ supply limitation were causing the decline in inactive and routine rates of metabolism they observed, then the much higher maximum metabolic rates documented should not have been attainable.

Other cases of lowered metabolic rate independent of O₂ supply limitation have been documented in marine animals exposed to hypercapnia. Hu et al. (2014) found that after 1 week of exposure to 420 Pa (4130 μatm) CO₂, Sepioteuthis squid were significantly metabolically suppressed even though their blood remained fully pH-compensated. Rummer et al. (2013) have also found that fish exposed to high CO₂ can have suppressed resting Ṁ_O2, yet have enhanced maximal Ṁ_O2, which would be impossible if O₂ supply is insufficient to even sustain resting needs. In the marine worm Sipunculus nudus, hypercapnia can cause metabolic suppression through alteration in neuromodulator concentration independently of O₂ supply constraints (Reipschläger et al., 1997). Additionally, hypercapnia-induced metabolic suppression in corals has been associated with differential gene expression of metabolic pathways at the tissue level (Kaniewska et al., 2012) rather than limited O₂ delivery.

As has been demonstrated in other adult cephalopods (Gutowska et al., 2010a; Hu et al., 2014; Häfker, 2012), it is likely that the squid species examined here in high seawater P_CO2 compensated their blood pH by actively increasing []. Sepioteuthis lessoniana have been found to fully compensate for respiratory acidosis within 20 h of exposure to 420 Pa (4130 μatm) CO₂ (Hu et al., 2014). Similarly, when exposed to 600 Pa (5920 μatm) CO₂, cuttlefish blood pH is nearly fully compensated and hemocyanin–O₂ saturation is not compromised (Gutowska et al., 2010a). In addition to raising blood pH, increasing [] from 3 to 10 mmol l⁻¹ lowers free [Mg²⁺] by ≈1% owing to ion pairing. Although free Mg²⁺ is essential for proper hemocyanin function, such a small change has a negligible effect on hemocyanin P₅₀ (Miller, 1985; Miller and Mangum, 1988).

In addition, hypoxia may even have an antagonistic effect of hypercapnia on blood pH. Seibel et al. (2014) reported that D. gigas blood pH increased under hypoxia, presumably to increase O₂ affinity although at the expense of intracellular pH. Similar blood pH increases in response to hypoxia have been measured in Octopus and Sepia (Houlihan et al., 1982; Johansen et al., 1982).

To determine what effect P_CO2 might reasonably be expected to have on O₂ supply capacity, we modeled the effect of P_CO2 across a range of physiological and temperature conditions. The model matches independent empirical data up to 3200 Pa or 31,500 μatm CO₂ (Redfield and Goodkind, 1929), which suggests that despite its simplicity, our model accurately captures whole-animal metabolic responses to CO₂ in squid when no blood acid–base compensation occurs. The model for D. pealeii (the more CO₂ sensitive of the two species) results in only a 0.52 kPa increase in P_crit owing to a rise in P_CO2 to 100 Pa (1000 μatm). This increase in P_crit is well within the range of existing intraspecific P_crit variability measured in this study and others (Redfield and Goodkind, 1929; Trueblood and Seibel, 2013).

We have been quite conservative in this analysis by intentionally constructing a model that does not incorporate a number of physiological phenomena that would further minimize the effect of OA on blood O₂ supply. There has been no incorporation of pH compensation via branchial ion transport. In fact, such compensation has been demonstrated in squid, and would greatly alleviate impacts of increased CO₂ (Melzner et al., 2009). Hu et al. (2014) found that the loliginid squid Sepioteuthis lessoniana exposed to 400 Pa (4000 μatm) CO₂ could raise blood pH by at least 0.13 units to completely return blood pH to normocapnic levels. Similarly, Gutowska et al. (2010a) found that Sepia officinalis could raise blood pH by ≈0.2 units in response to hypercapnia. Such compensation would completely relieve the effects of CO₂ on O₂ supply demonstrated here (Fig. 5A).

Some species of fishes, crustaceans and mollusks (including cephalopods) are known to produce multiple isoforms of respiratory protein subunits with different pH sensitivities (Johansen and Weber, 1976; Mangum, 1997; Strobel et al., 2012b). This could allow an animal to utilize a pH-insensitive isoform to further minimize impairment of O₂ supply. Such a response seems to occur in rainbow trout exposed to very high CO₂ (>1300 Pa, 13,000 μatm; Eddy and Morgan, 1969). In addition, a minor fall in arterial saturation may be compensated for by a slight increase in cardiac output, which would be particularly advantageous at high blood P_O2. Increases in cardiac output capacity have been observed in hypercapnia-exposed fishes (Gräns et al., 2014).

Water-breathers are also able to increase branchial surface area or decrease diffusion distance under projected OA-level hypercapnia (Esbaugh et al., 2016), which could lessen the increase in blood P_CO2. Increased ventilation of gas-exchange structures (e.g. gills) would also lower arterial P_CO2 and raise blood pH, and has been documented in cephalopods and fishes exposed to hypercapnia (Gutowska et al., 2010a; Ern and Esbaugh, 2016). Cephalopods, fishes and crustaceans have all been shown to modulate their hypoxia sensitivity between populations that inhabit different O₂ conditions, demonstrating that plasticity can further ameliorate environmental stressors to O₂ supply (Childress, 1975; Friedman et al., 2012; Birk, 2018).

Based on the small to modest effects of CO₂ on O₂ supply demonstrated here without blood acid–base compensation and the well-established abilities of squid to regulate blood pH, we propose that O₂ supply capacity in loliginid and ommastrephid squid is highly unlikely to be impaired to any great extent by OA predicted for the near future, even in hypoxic–hypercapnic environments. Although some squid may have slightly larger Bohr coefficients or higher P₅₀, these species have been held up as models of CO₂-intolerant species. We show that they are highly tolerant of likely OA scenarios even without inclusion of acid–base regulation capacity. This suggests that whole-animal metrics such as hypoxia tolerance (P_crit), maximal metabolic rate and aerobic scope are also unlikely to be impacted by OA in active squid.

Although it is unlikely that the O₂ supply pathway of active squid will be affected by OA, there remain other mechanisms of concern for hypercapnia to impact squid fitness. Although OA does not raise overall O₂ demand, the possibility remains that OA may alter the allocation of metabolic energy as a result of changes in the relative demands of the metabolic budget components (Strobel et al., 2012a; Pan et al., 2015). Blood acid–base disturbance from environmental hypercapnia has been shown to increase cuttlebone calcification in cuttlefish (Gutowska et al., 2010b). OA has also been demonstrated to alter behavior in marine animals (Clements and Hunt, 2015), including squid (Spady et al., 2014, 2018). Additionally, OA has been found to have negative impacts on embryonic growth rates and hatching success (Zakroff, 2013; Kaplan et al., 2013).

We found that ventilation rate increases under hypoxia in the loliginid squid D. pealeii, which increases the flux of O₂ past the gills as the O₂ content of the seawater declines. Similar responses have been observed in other cephalopods and aquatic animals generally (Wells and Wells, 1995; Hughes, 1973; Burnett and Stickle, 2001). However, this finding contrasts with results obtained in studies of another loliginid squid, Lolliguncula brevis, in which progressive hypoxia had no effect on ventilation rate but, instead, animals increased O₂ extraction efficiency (Wells et al., 1988). Extraction efficiency was not measured in the present study, so it is unclear whether D. pealeii exhibit a similar response in addition to their increased ventilation rate. Although ventilation was not measured in D. gigas here, Trübenbach et al. (2013) examined ventilation rate and stroke volume under normoxia and severe hypoxia (1 kPa O₂). At this extreme hypoxic condition (below P_crit), animals suppress metabolism. Thus, both ventilation rate and stroke volume are lower under extreme hypoxia than normoxia. It is still unclear, however, what effect intermediate P_O2 levels have on ventilation in this species before it begins to suppress metabolism. Pelagic crustaceans that also migrate daily into the OMZ have been found to increase ventilation with progressive hypoxia (Childress, 1971; Seibel et al., 2018).

CO₂ had no effect on ventilation rate (Fig. 3B). If blood acid–base balance can be fully compensated for by branchial ion-transport under elevated CO₂, then no increase in ventilation rate is necessary to maintain O₂ supply. This is common in water-breathing animals, which rely much less on respiratory adjustments for acid–base balance than air breathers (Pörtner et al., 2011). However, cuttlefish and some fishes have been found to increase ventilatory dynamics under hypercapnia (Gutowska et al., 2010a; Ern and Esbaugh, 2016). In fishes, this is driven not only by blood acidosis but also by CO₂-sensitive chemoreceptors in the gills (Gilmour, 2001). It is not currently known whether cephalopods also have such branchial chemoreceptors.

After adjusting for temperature, both species of squid had similar metabolic rates. Although the two species are phylogenetically rather distant (different orders), they are both active squid that inhabit shallow water and thus they both have strong selection for high metabolic rates (Seibel, 2007).

In this study, we found that the ommastrephid D. gigas has a better tolerance to hypoxia than the loliginid D. pealeii when compared at the same temperature. The hypoxia tolerance of aquatic animals can closely define their distribution and suitable habitat on spatial scales from meters (Mandic et al., 2009) to hundreds of kilometers (Deutsch et al., 2015). It is therefore unsurprising that D. gigas has better hypoxia tolerance than D. pealeii because the former is closely associated with the strong OMZ of the eastern tropical Pacific (Nigmatullin et al., 2001) whereas the latter is not known to frequently encounter such extreme hypoxia. Although the bays that D. pealeii inhabit in the northern part of their range may occasionally become hypoxic (Melrose et al., 2007), it is likely that this species can easily find suitable habitat outside these small spatiotemporal regions (Bartol et al., 2002) and thus minimize the selective pressure to improve hypoxia tolerance.

Dosidicus gigas are demonstrably better equipped to handle hypoxia than D. pealeii. The relative gill length of Dosidicus gigas is 30% longer than that of D. pealeii, which suggests greater gill surface area. Dosidicus gigas hemocyanin requires less than half the blood P_O2 to saturate its hemocyanin as D. pealeii (Pörtner, 1990; Seibel, 2013). Finally, D. gigas maintain far higher anaerobic capacity than D. pealeii. Dosidicus gigas store 2–4× higher concentration of phosphoarginine in its mantle muscle than D. pealeii (Seibel et al., 2014; Storey and Storey, 1978), which should be more advantageous for surviving subcritical O₂ levels. Glycogen reserves have not been quantified in D. pealeii mantle muscle, but glycogen concentration in D. gigas mantle is ≈300 μmol glucosyl units g⁻¹ (Seibel et al., 2014), which is much higher than in most fishes (Nilsson and Östlund-Nilsson, 2008; Richards, 2009) and even in bivalves that can survive months in anoxia (Oeschger, 1990).

In this experiment, we found that P_CO2 levels up to 122 Pa (1200 μatm), near the P_CO2 in the OMZ (Paulmier et al., 2011; Feely et al., 2016), had no measurable effect on D. gigas P_crit. Therefore, we do not expect that hypercapnia encountered in the OMZ has any additive or synergistic effect with hypoxia on D. gigas during its daily vertical migrations into this region.

In conclusion, although shallow-water squid species have high O₂ demand and constrained O₂ supply, their blood-O₂ carrying capacity, hypoxia tolerance and O₂ demand seem to be unaffected by near-future CO₂ levels.

We would like to thank Ed Baker, Amy Maas, Dennis Graham, Ann Kelly, Jillon McGreal, Joe Langan, Xuewu Liu and the crew of the R/V Oceanus for methodological support, as well as Agnieszka Dymowska, Tracy Shaw, Alyssa Andres, Yue Jin and two anonymous reviewers for helpful comments on the manuscript.