## Introduction

Increasing anthropogenic carbon dioxide (CO2) emissions are causing atmospheric CO2 concentrations to rise at a rate unprecedented for millions of years1. The global ocean acts as a buffer for rising atmospheric CO2 levels, as CO2 is sequestered in the surface waters. This absorption of CO2 at the air-ocean interface makes seawater more acidic (ocean acidification), due to an increase in the partial pressure of carbon dioxide (pCO2), hydrogen ions and carbonic acid in seawater2. The atmospheric CO2 concentration has increased by 120 μatm since the industrial revolution (ca. 1850), causing a 0.1 pH unit drop in ocean surface waters3,4. Model projections suggest that if anthropogenic emissions are not reduced this will result in a further decrease of 0.3–0.5 pH units by the year 2100, and 0.77 units by 23005,6.

Ocean acidification has negative effects on some marine organisms, causing decreased mineralisation or dissolution of calcium carbonate shells, decreased or delayed growth, increased mortality and delayed reproduction or abnormalities in offspring7. Ocean acidification also causes an increase in pCO2 (and decrease in pH) in the intra- and extra- cellular spaces of marine organisms, as CO2 diffuses across cell membranes8,9,10. The acid–base balance of extracellular fluids must be kept within a certain range for animals to carry out important biochemical functions11, prevent metabolic depression and transport oxygen around the body8. Despite this range of negative effects, animal responses to acidification are species-specific and a range of positive, negative and neutral responses have been observed in organisms exposed to increased seawater pCO2 in the laboratory8,12. Active crustaceans may be more resilient to ocean acidification than other taxonomic groups, due to their increased ability to regulate extracellular pH (pHe) compared with more sessile taxa9,11.

Euphausia superba (Antarctic krill, hereafter krill) is the primary prey of marine mammals, penguins and seabirds in the Southern Ocean, which makes it a keystone species in this region13. Krill are also the target of the region’s largest fishery14. They are highly active crustaceans, and their ability to exploit their environment makes them one of the most abundant organisms on Earth15.

The Southern Ocean is a major carbon sink16 and predictions suggest that ocean acidification will occur most rapidly in this region17. Seawater pH in the Southern Ocean varies with season (pH is lower in winter than summer18), and pCO2 is highest at intermediate depths19,20. Early life stages of krill (eggs, embryos and larvae) sink to 700–1000 m depths during their development before migrating back to surface waters21, and adult krill have been found as deep as 3500 m22. Therefore, they may already be exposed to pCO2 levels up to 550 μatm during their life cycle20. Model projections have shown that the Weddell Sea may reach 1000 μatm pCO2 at the surface, and 2000 μatm pCO2 at depth, within the next 80 years19.

Previous short-term studies indicate that Antarctic krill may be more vulnerable to ocean acidification than crustaceans from lower latitudes. Krill eggs fail to hatch at pCO2 levels predicted to occur by the year 230019,20, adults increase feeding and nutrient excretion at 750 μatm pCO223, and krill may not have the behavioural ability to discriminate between low pCO2 and high pCO2 seawater24.

Understanding how organisms will respond to high CO2 requires laboratory experiments that measure a wide range of physiological performance indicators over periods of months or years9,11,25. To our knowledge, we conducted the first long-term laboratory study to investigate the effects of ocean acidification on adult Antarctic krill. Adult krill were reared for a 46-week period that encompassed all four seasons (25th January – 12th December 2016). Krill were reared in present day seawater pCO2 concentrations (400 μatm pCO2, the control), a range of seawater pCO2 levels predicted to occur in their habitat within the next 100–300 years (1000–2000 μatm pCO2), and an extreme level of 4000 μatm pCO2. Throughout the 46-week experiment we measured a suite of physiological and biochemical variables, to investigate how future ocean acidification may affect the survival, size (total length), lipid stores (triacylglycerol), reproduction (maturity and female ovarian development), metabolism (respiration rate) and extracellular fluid (haemolymph pH) of krill. We show that these physiological processes in krill are largely unaffected by pCO2 levels predicted within the next 100–300 years. Adult krill are able to actively maintain their extracellular pH in 400–2000 μatm pCO2, which enhances their resilience to ocean acidification.

## Results

### Survival

The survival rate of krill was highest in present day and future pCO2 seawater (400–2000 μatm) throughout most of the experiment (Fig. 1). The survival rate of krill by week 46 was higher in the 1000–2000 μatm pCO2 treatments (87–90%) than the control (400 μatm) treatment (79%). Large decreases in survival rate occurred between weeks 3–7 and weeks 19–22 in the extreme pCO2 (4000 μatm) treatment and plateaued towards the end of the experimental period (weeks 29–46), with 53% of individuals surviving by week 46 (Fig. 1).

### Body length and triacylglycerol content

Krill in all treatments maintained their total length and triacylglycerol content (fat stores) during summer (weeks 1, 2, 4 and 5; Fig. 2 and 3), with no differences observed between pCO2 treatments or weeks for length (Two Way ANOVA, pCO2*week; df = 12, F = 1.12, p = 0.359) or triacylglycerol content (Two Way ANOVA, pCO2*week; df = 12, F = 1.14, p = 0.341).

By winter (week 26), the median length of krill in all treatments had decreased (Fig. 2). Krill in extreme pCO2 seawater (4000 μatm) were significantly shorter (Dunnett test, p = 0.023) and had stored less fat (Dunnett test, p = 0.041) than krill in ambient pCO2 seawater.

Throughout spring, krill in the 4000 μatm pCO2 treatment were shorter than krill in 400 μatm pCO2 (Dunnett tests week 39; p= 0.094, week 41; p< 0.001, week 43; p= 0.005), but no differences were seen between treatments by the following early summer (week 46; One Way ANOVA pCO2; df = 4, F = 0.73, p= 0.584). Triacylglycerols remained lower in krill in the 4000 μatm pCO2 treatment compared with krill in ambient pCO2 throughout early spring (Dunnett tests week 39; p = 0.021 and week 41; p < 0.001), but all treatments had similar triacylglycerol content by late spring (week 43; One Way ANOVA pCO2; df = 4, F = 1.13, p = 0.379).

### Sexual maturation and ovarian development

The sexual maturity of krill in the 400–2000 μatm pCO2 treatments advanced between spring and early summer (weeks 39–46) and all krill reached maturity at similar times (Supplementary Table 1). When maturity scores from weeks 39–46 were combined, overall maturity scores of krill were lowest in the 4000 μatm pCO2 treatment, suggesting delayed sexual development (Fig. 4a). Krill in 400–2000 μatm pCO2 had completed ovarian development to the previtellogenesis or early vitellogenesis stages by week 46, but ovarian development in krill in the 4000 μatm pCO2 treatment was delayed and had not progressed past oogenesis (Fig. 4b).

### Respiration rate

Respiration rates of krill in early spring (week 38) ranged from 0.13–0.50 μL O2 mg DM h−1 (Fig. 5) and did not differ between pCO2 treatments (One Way ANOVA, pCO2; df = 4, F = 1.26, p = 0.301). Intraspecific variation in the respiration rates of individual krill increased in elevated pCO2 treatments (Fig. 5).

### Haemolymph pH

Haemolymph pH of krill measured in week 46 ranged from pH 7.57–8.47. Haemolymph pH of krill in 1000–2000 μatm pCO2 treatments did not differ significantly from the control (Dunnett tests; 1000 μatm p= 1.000, 1500 μatm p= 0.145, 2000 μatm p= 0.369) (Fig. 6). The average haemolymph pH of krill in the 4000 μatm pCO2 treatment was 0.5 units lower than krill in the control treatment (Dunnett test, p < 0.001). There was a linear trend of decreasing haemolymph pH with increasing pCO2 (One Way ANOVA with polynomial contrasts, pCO2; df = 4, F = 11.69, linear p < 0.001).

## Discussion

Our experimental results show that the measured physiological processes in adult Antarctic krill were robust to near-future ocean acidification (1000–2000 μatm pCO2), when elevated pCO2 was assessed as a single stressor. The survival rate of krill subject to near-future pCO2 increased by up to 11%, and seasonal patterns of growth, fat storage and reproductive development were comparable to wild krill26,27,28. These physiological processes appeared to be controlled by endogenous rhythms25,29,30, and were not affected by near-future pCO2.

Most studies report a decrease in survival when organisms are exposed to acidification7. In contrast, slight increases in euphausiid survival rates have been observed in Euphausia pacifica after a 2-month exposure to 1200 μatm pCO231 and in Nyctiphanes couchii after a 35-day exposure to 800 μatm pCO2 seawater32. Euphausiids that are exposed to vertically changing pCO2 in the water column may use acid-base regulation and short-term metabolic depression (reduced respiration rates) to enhance survival in high pCO2 conditions31,33.

Primary productivity may increase in high pCO2 seawater, increasing food supply and subsequent survival of herbivores in these experimental treatments34. It is unlikely that phytoplankton growth (and therefore food supply) increased in our high pCO2 tanks, as phytoplankton was grazed by krill within ~2 h. Furthermore, the majority of phytoplankton added to tanks were non-viable cultures that do not photosynthesise. Further targeted studies on krill survival under ocean acidification conditions, and effects of pCO2 on their food sources, may identify whether krill survival is enhanced in elevated pCO2 seawater.

In our study, pCO2 levels between 1000–2000 μatm did not affect the size of adult krill over a whole year and this reflects their ability to moult and grow. Reduced growth rates have been observed in adult crustaceans exposed to high pCO2 seawater for short-medium term durations (weeks to months)11. Elevated pCO2 did not affect growth rates in the north Atlantic euphausiids N. couchii32 or Thysanoessa inermis35 after short-term (5–11 week) exposure, but exposure to levels of 1200 μatm pCO2 over 2 months slowed growth in E. pacifica31.

Krill in our study were shorter than wild krill which can grow up to 60 mm in total length36. Growth of wild krill is closely related to food quality and quantity, and laboratory reared krill do not grow as large as wild krill37. It is impossible to directly replicate the wild diet in controlled conditions, therefore the shorter lengths attained by krill in our study may have been caused by lower food quality in the laboratory. Patterns of seasonal growth seen in wild krill (e.g. winter shrinkage) were, however, observed in our experimental krill, suggesting that the experimental conditions replicated the physiological cycle of wild krill as closely as possible.

The resilience of Antarctic krill, in terms of their maturation and ovarian development to near-future pCO2, is comparable to other pelagic crustaceans. Short-term studies (<2 weeks) have generally found that egg production is not affected by moderately increased pCO2 levels38,39, but production rates decrease significantly in crustaceans exposed to extreme pCO2 levels40.

Decreased growth and delayed reproduction are often observed in sessile organisms that cannot maintain their acid-base balance and those that decrease their metabolism when exposed to high pCO27. This occurs because energy is diverted away from growth and reproduction, and prioritised for acid-base compensation11. The ability of active Antarctic krill to maintain their size and mature in 1000–2000 μatm pCO2 is likely to be directly linked to their ability to maintain acid-base balance and respiration rates at these pCO2 levels.

An increase in krill metabolic activity has been observed after short term (24 h) exposure to ocean acidification23, suggesting that krill may raise their respiration rate on initial exposure to high pCO2. The increasing variation in krill respiration rates at higher pCO2 levels suggests that individuals vary in their capacity to respond to CO2-induced metabolic stress. This may be due to intraspecific differences in phenotypic plasticity, or genetic predisposition to metabolic resilience in some individuals41,42.

The ability of krill to maintain their acid-base balance in elevated pCO2 seawater may be the key to their successful survival, maturity and growth in a future high CO2 world. Haemolymph pH can be increased in hypercapnic conditions via ion transport pumps that pump bicarbonate into the extracellular space9,43. These pumps are located in the gill epithelia and consume energy as they actively transport ions in and out of body compartments8. Our results suggest that krill in elevated pCO2 were actively maintaining haemolymph pH, as it remained within the same range for krill in 400–2000 μatm pCO2. The negligible effects on growth and reproduction in these krill indicate that they were able to actively regulate acid-base balance at low energetic cost. The trend of decreasing haemolymph pH with increasing pCO2 indicates that although krill in near-future pCO2 were able to maintain haemolymph pH within the same range as krill in ambient pCO2, measurements were within the lower range of values for krill in ambient pCO2. This may have implications for longer term acid-base maintenance. The ability for krill to maintain haemolymph pH beyond one year, and into their spawning season, is unknown. The substantial increase in mortality in extreme pCO2 (4000 μatm) may have been caused by the inability of those krill to maintain acid-base balance.

Unlike decapods with gills located inside their carapace, Antarctic krill have external gills with a complex structure built for efficient ion and gas exchange44. These intricate gills are designed to maximise the amount of O2 available to krill during swarming and frequent periods of intense swimming activity44. The ability to rapidly exchange O2, CO2, and ions across their external gills may have assisted krill in maintaining acid-base balance and respiration rates when exposed to near-future acidification. Modification of the respiratory pigment haemocyanin may also assist crustaceans to maintain levels of O2 consumption during hypercapnia11, however this was not measured in our study.

Krill have evolved a unique range of adaptations to survive the Antarctic winter29. Metabolic depression is one such adaptation that is controlled endogenously, cued by the seasonal light cycle26,45,46. The physiological responses of krill in our extreme pCO2 treatment (4000 μatm pCO2) suggest that this energy-conserving strategy may be less advantageous in high pCO2 conditions. In winter, as the photoperiod approached 24-h darkness, krill growth and fat deposition in 4000 μatm pCO2 seawater were reduced compared with krill in ambient pCO2. In this extreme environment, metabolic depression during winter may have prevented krill from maintaining respiration rates high enough to maintain pHe, grow and store fat. These reductions in winter growth and fat storage may have contributed to the subsequent delay in reproductive development47.

The energy needed to maintain pHe can be met by consuming more food48, and Antarctic krill do increase their feeding rates in elevated pCO2 seawater23. The constant food supply in our experiment may have enabled krill in the 4000 μatm pCO2 treatment to perform better than if they had received food at seasonally variable concentrations. Importantly, this may have also enabled krill in lower pCO2 treatments (1000–2000 μatm pCO2) to maintain haemolymph pH, normal growth, and reproductive development. The relationship between food supply and pCO2 can affect predator physiology in different ways49, and requires further investigation. Metabolic depression, the increasing severity of winter acidification18, and regionally variable food concentrations50 may increase the vulnerability of krill to near-future ocean acidification during winter.

The prosperity of Antarctic krill in a high CO2 world will depend on the ability of adults to produce offspring resilient to ocean acidification. If early life stages cannot survive, this may have catastrophic consequences for krill populations and the Southern Ocean ecosystem. Previous studies indicate that krill eggs and embryos are sensitive to seawater pCO2 above 1250 μatm19,20. These studies used gametes from parents that were maintained in ambient pCO2 conditions, and gametes were spawned into ambient seawater before being subjected to high pCO2 conditions. Recent studies have shown that some adult echinoderms and molluscs that acclimate to high pCO2 conditions are able to produce gametes resilient to high pCO251,52, and this may allow such species to adapt to ocean acidification over generational time scales53. Further studies may establish whether this generational adaptation occurs in krill, which would influence the way that we assess the vulnerability of the early life stages.

Our results suggest that adult Antarctic krill are resilient to ocean acidification, and may not be affected by pCO2 levels predicted for the next 100–300 years. The overall resilience of Antarctic krill as a species will, however, depend on long-term effects occurring at all life history stages. Endogenous rhythms controlling metabolic rate, combined with food availability in the wild, may influence the vulnerability of krill to high pCO2 in winter. Negative effects on krill physiology may be seen at near-future pCO2 levels if effects of acidification are exacerbated by other stressors such as ocean warming. The persistence of krill in the Southern Ocean is vital for the health of the Antarctic ecosystem, and we are only just beginning to understand how this keystone species may respond to climate change.

## Methods

### Experimental conditions

Live krill were collected on the RSV Aurora Australis via rectangular mid-water trawl on 22–23 February 2015 (66–03°S, 59–25°E and 66–33°S, 59–35°E). Krill were held in shipboard aquaria using standard maintenance methods54 before being transferred to the Australian Antarctic Division’s (AAD) Krill Aquarium in Tasmania (seawater temperature 0.5 °C and pH 8.1). Seawater was supplied to aquarium tanks via a seawater recirculating system55.

For ocean acidification experiments, 0.5 °C seawater was supplied from a 70 L header tank and equilibrated with air (control) or CO2-enriched air (elevated pCO2 treatments) before delivery to experimental tanks. The CO2-enriched air was monitored using mass flow controllers (Horiba STEC SEC-E-40) and air valves, to regulate flow rates of atmospheric air and pure CO2. Five experimental 300 L tanks were maintained at five pCO2 levels; control 400 μatm pCO2 (pH 8.1), 1000 μatm pCO2 (pH 7.8), 1500 μatm pCO2 (pH 7.6), 2000 μatm pCO2 (pH 7.4) and 4000 μatm pCO2 (pH 7.1).

Appropriate tank size and the best possible animal husbandry were high priorities in such a long-term study. As krill are a pelagic species, large sized (300 L) experimental tanks were needed to emulate wild conditions as closely as possible in a laboratory. Our experimental design was limited by the space and resources needed for these large tanks, and our observational units (CO2 treatment tanks) could not be replicated. We did not however, observe any visual evidence to suggest that tank effects were confounding our results.

Two hundred krill were randomly assigned to each experimental tank on 25 January 2016, corresponding to a density of 0.6 individuals L−1. This density is in the range of 0.5–2 individuals L−1 which has been successfully used in previous experiments at the AAD krill aquarium30,45. The experiment ran for 46 weeks from the 25 Jan 2016–12 Dec 2016 covering all four seasons. Mortality rates in all pCO2 treatments (ranging from 0.03–0.2% day−1) were much lower than previously reported for Antarctic krill in shipboard aquaria (2% day−1)54 and in other pCO2 studies on Pacific krill (0.5% day−1)31 and northern Atlantic krill (5% day−1)32.

The pCO2 levels of the CO2-enriched air and seawater were monitored daily using a LI820 CO2 gas analyzer and associated computer software (version 2.0.0), and daily pH levels of experimental tanks were measured manually using a pH meter (Mettler Toledo SevenGo Duo Pro). A three-point calibration of the pH meter was undertaken daily using Radiometer Analytical IUPAC Standard buffers of pH 7.000, 7.413 and 9.180. Total alkalinity (AT) and dissolved inorganic carbon (DIC) were measured weekly using a Kimoto ATT-05 Total Alkalinity Titrator. Salinity was measured weekly using a Profiline™ Cond 197i Conductivity Meter, WTW. The average total pH (pHT), pCO2, calcite and aragonite saturation (ΩC and ΩA) values over the 46 week experiment were calculated in CO2SYS56 using our measured salinity, temperature, alkalinity and DIC data, and using equilibrium constants of Merhbach, as modified by Dickson and Millero57. Average levels of pCO2 were 8–169 μatm below target levels for the 400–2000 μatm treatments, and 123 μatm above the target level for the 4000 μatm treatment. Seawater temperature and AT were stable in all treatments, while DIC increased with increasing pCO2. Seawater chemistry in the experimental aquarium is shown in detail in Supplementary Table 2.

Krill were fed 6 days per week with a mixed microalgal diet of the Antarctic flagellate Pyramimonas gelidicola at a final concentration of 2 × 104 cells mL−1, and Reed Mariculture Inc. (USA) cultures of the diatom Thalassiosira weissflogii (8.8 × 103 cells mL−1), flagellate Pavlova lutheri (4.5 × 104 cells mL−1) and flagellate Isochrysis galbana (5.5 × 104 cells mL1)30,37.

Light was controlled in the laboratory to ensure that the photoperiod mimicked the seasonally changing light regime of the Southern Ocean (66°S, 30 m depth). Photoperiod was altered monthly, with a maximum of 100 lux light intensity in February and minimum intensity (24 h darkness) in August (Supplementary Table 3). Light was provided by twin fluorescent tubes and was controlled via standard aquarium procedures55.

### Survival

Each pCO2 treatment was checked daily for mortalities, which were recorded and placed in vials of 10% formalin. Daily mortality data were used to calculate the percentage of krill still surviving at the end of each experimental week in each treatment using the equation:

$${\%}\,{\rm{krill}}\,{\rm{remaining}}\,{\rm{prev}}\,{\rm{week}}-\frac{{{\rm{Num}}\,{\rm{mortalities}}\,{\rm{current}}\,{\rm{week}}}}{{{\rm{Number}}\,{\rm{of}}\,{\rm{krill}}\,{\rm{remaining}}\,{\rm{in}}\,{\rm{tank}}}}\ \times 100$$

where "% krill remaining prev week" is the percentage of krill remaining in the previous week and "Num mortalities current week" is the number of mortalities during the current week. Krill that were sampled for experimental purposes were not counted as mortalities, but were subtracted from the number of krill remaining in the tank each week. This ensured that the remaining number of krill used to calculate survival percentages reflected actual experimental mortality.

### Total length

Krill lengths (mm) were obtained from krill in each pCO2 treatment in weeks 1, 2, 4, 5, 26, 39, 41, 43 and 46. Sample sizes (n) for length measurements for each week and treatment are shown in Supplementary Table 4a. Individuals were sexed using microscopy and the length of each specimen was measured from the tip of the rostrum to the tip of the uropod (measurement Standard Length 158). Length data from frozen krill and live krill were combined.

### Lipid class analysis (triacylglycerols)

Lipid analysis focused on triacylglycerols which are the main storage fat in krill and, therefore, drive overall lipid concentrations and lipid class composition of krill27. Krill were sampled for lipid analysis from all pCO2 treatments in weeks 1, 2, 4, 5, 26, 39, 41 and 43. Individual krill were placed in cryo-tubes and immediately stored in a –80 °C freezer.

Lipid class analysis was carried out on 4-5 krill from pCO2 treatments 400, 1000, 1500 and 2000 on each sampling week (n = 3 for the 4000 μatm tank in weeks 39, 41, and 43 due to increased mortality and lower numbers of krill in that treatment). Sample sizes (n) for each week and treatment are shown in Supplementary Table 4b. The wet mass (g), total length (measurement Standard Length 158), and sex for each krill was obtained, and krill were kept frozen during this process to prevent sample degradation. A dry mass (g DM) was obtained later by multiplying the wet mass by 0.2278 to account for the 77.2% water content in the organism59. Total lipid extracts of krill specimens were obtained using a modified Bligh and Dyer method60,61. Lipid class composition and content were determined using an Iatroscan MK-5 TLC/FID Analyser using standard methods27.

### Sexual maturation

The maturity stages of individual krill were identified during weeks 39, 41, 43 and 46 (n = 5 for 400–2000 μatm pCO2 treatments, n = 3 for the 4000 μatm pCO2 treatment). Adult krill undergo sexual regression in winter, so these measurements occurred at the end of the experiment to capture the onset of maturity during late spring/early summer.

The sex and maturity stage of each krill was identified via microscopy (using the staging key in Supplementary Table 5). Each maturity stage was given a maturity score with higher numbers denoting greater maturity (Supplementary Table 5). After staging, individual krill were placed in a cryopreservation tube with 10% formalin.

### Ovarian development

On the final day of the experiment (12 December 2016, Week 46), krill left in each experimental tank were preserved in 10% formalin. These samples were used to determine the ovarian development of eight randomly selected females from each of the 400, 1000, 1500 and 2000 μatm pCO2 treatments. Only two females remained in the 4000 μatm pCO2 treatment, therefore only two replicates could be obtained for this tank.

The ovary was dissected out of each organism and a single lobe was placed on a microscope slide with a drop of distilled and deionized water and lightly squashed62. Photographs were taken of the ovary section and the lengths of the largest cells (across the longest axis of the cell) were measured using the computer software Image J (https://imagej.nih.gov/ij/). The cell size and photographs were used to determine the maturation stage of krill ovaries using the key in Supplementary Table 6 (modified from Cuzin-Roudy & Amsler62). When an ovary was transitioning from one stage to another, a 0.5 value was used (e.g. 4.5). Photographic examples of different ovary stages are shown in Supplementary Figure 1.

### Respiration rate

Respirometry measurements were carried out in experimental week 38. Respirometry vessels (2 L) with pre-fitted O2 mini sensors were filled with seawater sourced from the inlet hose of each experimental tank and placed in a 0 °C water bath. Each vessel was connected to O2 computer software (version OXY10v3_50TX) via an optic fibre probe.

Ten krill were sampled from each experimental tank (n = 50 total) and the total length (measurement Standard Length 158) and wet mass (g) were obtained for each individual. A dry mass (g DM) was obtained by multiplying the wet mass by 0.2278 to account for the 77.2% water content in the organism59.

Each krill was then placed into a respirometry vessel completely filled with experimental seawater, with no air spaces in the vessels. Oxygen saturation (%) was logged at 5 min intervals in each respirometry vessel over 22 hrs (9AM–7AM the following day), using the computer software. The software was calibrated at 0 °C and the atmospheric pressure at the time of measurement. After 22-hrs krill were removed from the vessels and returned to their experimental tanks.

Only measurements of O2 saturation (%) taken between 12PM–7AM were considered for analysis, to ensure that krill had three hours at the beginning of respiration trials to settle into a normal rhythm of respiration before data was collected. Oxygen saturation (100 %) for seawater at 0 °C and 35.1 salinity units (‰) was converted to O2 mL L1 using the equation in Fox63 to obtain a value of 8.035 mL L−1. This was used to convert the O2 saturation (%) at each logged time point to millilitres of O2 (O2 mL) in each 2 L respirometry vessel using the equation:

$${\mathrm{O}}_2\;{\mathrm{ml}}\;{\mathrm{in}}\;{\mathrm{respirometry}}\;{\mathrm{vessel}} = \frac{{\% \ {\mathrm{O}}_2\;{\mathrm{saturation}}}}{{100}} \times \left( {8.035 \times 2} \right)$$

Values for O2 mL in each respirometry vessel between 12PM and 7AM were used to create regression equations which were used to compute the O2 ml used in each respirometry vessel during this period. This value was divided by the krill dry mass (in mg), converted to µL O2 mg DM1, then divided by 19 h to obtain the µL O2 mg DM hr−1.

### Haemolymph pH

The haemolymph pH of five krill from each experimental tank was measured in week 46. Haemolymph pH was measured in situ by inserting a pH microelectrode directly into the pericardial cavity. This ensured that air contact with the haemolymph was minimised, as contact with air may alter the CO2 concentration and pH of the body fluids64. A Unisense pH Microelectrode (model pH-50, tip diameter 40–60 µm) and Unisense Reference Electrode connected to a Unisense pH/mV Metre and computer software (SensorTrace Logger) were used to complete measurements. The pH microelectrode and reference electrode were calibrated using the SensorTrace software via a three-point calibration using Radiometer Analytical IUPAC Standard pH buffers 7.000, 7.413 and 9.180. The buffers were chilled to the seawater temperature in which haemolymph measurements were conducted (0–0.5 °C). The pH of these buffers at 0 °C was used for calibrations (pH 7.12, 7.53 and 9.46 respectively).

Krill were individually removed from their 300 L tanks and placed under a compound microscope in a refrigerated microscope stage, submerged in seawater from the tank they originated from. The pH of this seawater was measured using the microprobe and reference probe, and a portable pH meter (Mettler Toledo SevenGo Duo Pro) to ensure that the measurements matched to within <0.05 pH units before proceeding.

Live krill were restrained within the microscope stage using acrylic blocks, designed to expose the integument that links the krill carapace to the abdomen. A micromanipulator was used to position the microelectrode relative to the animal. A camera connected to the compound microscope was also used to magnify the krill carapace-abdomen joint and view the real-time image on a computer monitor to ensure the accuracy of microelectrode placement.

The microelectrode was inserted through the integument underneath the carapace and into the pericardial cavity between the thorax and first abdominal segment. The reference probe remained in the seawater surrounding the krill during this process. Some resistance was observed as the microelectrode pierced the integument, causing a slight tear in the body wall as the probe penetrated the integument, ensuring that electrical conductivity was maintained between the reference probe and microelectrode.

The SensorTrace Logger software logged the pH of the haemolymph, and the pH was recorded once the reading had stabilised after ~1 min. The microelectrode was then withdrawn from the abdomen and haemolymph was observed leaking into the surrounding seawater as positive pressure from within the animal pushed it outwards. The krill was removed from the microscope stage and preserved in 10% formalin.

### Statistical analyses

Data were analyzed in the RStudio statistics package (version 0.99.893) using one-way ANOVA with pCO2 treatment as a factor, or two-way ANOVA with pCO2 and Week as factors. Dunnett comparisons (carried out using the RStudio multcomp package) were used to identify significant differences between the control treatment (400 μatm pCO2) and all other factor levels, while Tukey Post-hoc comparisons were used to compare all factor levels with one another. Polynomial contrasts were used to identify linear, quadratic and cubic trends in the data. Type 3 Sums of Squares (SS) were used when data was unbalanced and Type 1 SS were not appropriate. Data were log or square root transformed when assumptions of normality or homogeneity of variances were not met. For all analyses, α was set at 0.05 and all tests were two tailed. The RStudio packages ggplot2, plyr and dplyr were used to produce all figures.