Local differences in robustness to ocean acidification

Ocean Acidification (OA) results from increased anthropogenic atmospheric CO₂, which drives changes in the carbonate chemistry of ocean water, including the saturation state of aragonite (Ω_ar) and corresponding pH (Salisbury et al., 2008; Breitburg et al., 2015; Feely et al., 2010; Gledhill et al., 2015). This change in carbonate chemistry has been demonstrated to affect calcified organisms, especially early life stages of bivalved molluscs (Andersson et al., 2008; Feely et al., 2010; Gazeau, et al., 2013; Waldbusser and Salisbury, 2014; Waldbusser et al., 2015; Siedlecki et al., 2021; Vargas et al., 2022). Increased monitoring of ocean and coastal water chemistry has revealed that coastal areas are already experiencing greater OA conditions (termed coastal acidification) than those projected for the oceans in the next 50–100 years (Ekstrom, et al., 2015; Hunt et al., 2022). Because changes in ocean chemistry are occurring at least 100 times more rapidly than any changes experienced over the past 100,000 years (Zeebe, 2012; Zeebe et al., 2016), understanding the ability of species to respond or adapt is critical (Coen et al., 2007; Cranford, et al., 2007; Cranford, 2019; van der Schatte Olivier et al., 2020). Prior OA work with marine calcifying animals has generally found large negative effects of OA on survivorship, primarily in early life stages such as larvae (Kleypas et al., 2006; Pörtner and Farrell, 2008; Melzner et al., 2009; Schoener, 2011; Kelly and Hofmann, 2013; Munday et al., 2013; Meseck et al., 2021). For bivalve larvae it appears that the saturation state of aragonite (Ω_ar<1.5), rather than pH per se, impairs shell building (Waldbusser et al., 2015). Effects of OA are, however, expected to be profound for a variety of physiological and developmental processes of many taxa (reviewed by Kroeker et al., 2013; Bindoff et al., 2019). We are just beginning to assess the variety of physiological and developmental processes that may be influenced by OA (Waldbusser and Salisbury, 2014; Somero et al., 2016), as well as the capacity of species to adapt to these changes. Thus far, it is apparent that not all species, or developmental stages within a species, are inhibited or negatively affected by OA conditions (Gooding et al., 2009; Gazeau et al., 2013). In some cases, differences among individuals within a given species have been found, indicating variation in the capacity to respond to OA (Gazeau et al., 2013). Most studies to date, however, have tested the effects of OA on only a single population or source stock of a species.

The number of species for which physiological experimental studies have been conducted to document changes in metabolism, growth, calcification, survival, and immune response across life-history stages in responses to OA continues to grow, especially for marine bivalves (Gazeau et al., 2013; Gledhill et al., 2015; Clements and Hunt, 2017). Most OA studies, however, are of short duration and include a single source population, making them unlikely to capture the full potential for response to OA conditions of a species across its geographic range. Using these data to predict long-term effects on species abundance (Cooley et al., 2015; Rheuban et al., 2018; Grear et al., 2020), is challenging at best. In addition, most models projecting the effects of OA do not include the potential for adaptation or an evolutionary response to OA conditions within species (e.g. Cooley et al., 2015; Grear et al., 2020). Thus, the potential for species to develop resilience to increased levels of OA (acclimation or adaptation), especially differences among populations from different locales, remains largely unexplored.

Understanding the capacity of animals to respond to climate-driven changes in local conditions is critical to evaluate the capacity of species to respond to environmental change in general, including OA (Dupont et al., 2010; Pistevos et al., 2011; Sunday et al., 2011; Grear et al., 2020). Recently, studies of blue mussels, Mytilus edulis, from populations periodically naturally enriched with CO₂ found that their larvae had greater survivorship than those from populations without enrichment, showing short term selection and potential adaptation to CO₂ (Thomsen et al., 2017), and suggesting the potential for local adaptation to OA.

To understand the long-term effects of OA on bivalves in coastal ecosystems (Reusch, 2014; Thomsen et al., 2017), we need to determine whether there is capacity within species to adapt to these environmental changes. In addition, it is important to determine whether there are differences in response to OA conditions among populations (Harvey et al., 2016; Wood et al., 2016; Thomsen et al., 2017) from areas with different environmental conditions. To address these questions, we focused on the Mytilus edulis, an important model species for studies of physiology and adaptation (Goldberg, 1975; Menge, 2000; Meseck et al., 2021). Mytilus edulis are common to most temperate Atlantic shores, and have been a model species for studying physiological responses under different environmental conditions for decades (Bayne and Newell, 1983). They are also used as sentinel organisms to monitor coastal water for pollution (Farrington et al., 2016; Beyer et al., 2017). Blue mussels are an important foundation species in both the eastern and western north Atlantic and prolific suspension feeder that occurs in very high densities that can represent up to 90% of the coastal benthic biomass (Enderlein and Wahl, 2004). Blue mussels provide numerous ecosystem services, including improving water quality and stabilizing shorelines, among others (Nielsen et al., 2016; Smaal et al., 2019). The blue mussel is also a commercially important species in Europe and North America, valued at $0.7 billion annually (US dollars, http://www.fao.org/fishery/culturedspecies/Mytilus_edulis/en). Recent years, however, have seen a decline in blue mussel densities in the northeastern USA; scientists have suggested that much of this decline has been directly or indirectly attributed to effects of climate change, including OA (Sorte et al., 2017).

To address the question of the potential for adaptation to three different OA conditions, we conducted controlled experiments with larvae through metamorphosis of blue mussels from four populations sampled from Long Island Sound (Fig. 1) sites that differ in local water conditions (Table 1).

The effect of OA on survivorship of larvae was different among populations (Fig. 2). Both population (P=1.88e-238) and OA (P=0.000e+00) had highly significant effects on survivorship, and there was a significant population×OA interaction (P=9.78e-263; Table 2, Fig. 3), with populations responding differently to OA. Under the most extreme conditions (high OA) no larvae from any of the four populations survived to the end of the experiment, but larvae from Milford survived much longer than those from other populations. Larvae from the Stony Brook population had the lowest probability of survivorship; survivorship declined with increased OA and these larvae survived only in the low-OA treatment. Survivorship of blue mussel larvae from both Orient and Groton was not affected by mid- and low-OA conditions. Highest survivorship was for animals from Milford in the mid-OA treatment, which had much higher survivorship than those grown under low-OA conditions.

As observed for survivorship, there was a significant difference among populations in the rate of development as measured by time to metamorphosis (P=0.0002; Fig. 4 and 5, Table 2). Larvae from Groton had the fastest development, significantly faster than those from Orient, which were the slowest to reach metamorphosis. Contrary to survivorship, there was no effect of OA treatment on development time (P=0.6485), or OA treatment by population interaction (P=0.2597).

We found striking differences in robustness to OA conditions for blue mussels from the four different populations sampled from locations within Long Island Sound. For all populations, there was no survivorship to metamorphosis in the high-OA treatment, the most extreme OA condition tested. We did find differences in survivorship among populations for the mid-OA and low-OA conditions. Mussels from both Orient and Groton, the two easternmost sites, where water conditions presently are more similar to oceanic conditions. Mussels at these sites rarely experience the OA conditions tested in our experiments (Table 1), but were robust to lower pH, lower Ω_ar and higher pCO₂. There was no difference in survivorship between mid-OA and low-OA conditions for these two populations. For the other two populations, Stony Brook and Milford, we did see an effect of OA on survivorship, but not in the same direction. Larvae from Stony Brook showed the predicted response, and had the greatest sensitivity to OA; only larvae under the lowest-OA condition survived. Larvae from Milford, less than 50 km north of Stony Brook across Long Island Sound, had higher survivorship under mid-OA conditions than low-OA conditions, and thus appeared better able to cope with OA. Both Stony Brook and Milford mussels regularly experience seasonal OA conditions as severe as those tested in our experiment. For the Milford population, these results may indicate local adaptation to OA conditions. Overall, our results suggest that there is capacity for adaptation to OA, and we already see differences among populations in robustness to OA.

Contrary to other studies and general predictions (e.g. Talmage and Gobler, 2010; Meseck et al., 2021), there was no effect of OA on development time for any population; however, there was a difference in development time among populations. Although larvae from Groton and Orient were similarly robust in terms of survivorship to OA above the most severe conditions tested, they differed significantly in development time. Larvae from Groton developed faster than those from other populations, while larvae from the other eastern-most population, Orient, developed slower than others. These results suggest that genetic differences may underlie differences in response among populations, even those that are relatively close geographically, but any such differences would need to be tested with further research.

Our results demonstrate that the conclusions drawn from a given experimental study for a species may depend on the population or source stock used. Contrary to predictions that animals currently experiencing degraded habitats and low pH conditions would be more robust to OA, the two populations experiencing near oceanic conditions were the most robust to the tested OA conditions. At the same time, larvae from the Milford population had highest survivorship in the mid-OA treatment, and actually had lower survivorship under the low-OA (higher pH) treatment. These results, and the differences among populations in development time, suggest that aspects of our findings may be a result of genetic differences among populations and potentially local adaptation; however, this will require further study.

Prior research on blue mussels based on allozyme analyses found genetic differences among populations in eastern Long Island Sound (Hilbish, 1985; Koehn and Hilbish, 1987; Koehn, 1991). Eastern most populations had the same allele frequencies as open-coastal populations, but populations to the west (less than halfway between our more western test populations and those in the east), had different allozyme frequencies (Koehn and Hilbish, 1987; Koehn, 1991). The authors explained these patterns as evidence for selection for low salinity tolerance. Using demographic data, Hilbish (1985) suggested that selection for low-salinity tolerance was not at the larval stage, but was the result of differential juvenile survivorship. Therefore, a variety of environmental conditions may be affecting blue mussels at different stages during the life cycle. Determining whether there are differences in robustness to increasing OA conditions across life stages, and genetic potential for adaptation to OA, will require long-term experiments where individuals are exposed to increasing OA conditions across those life stage transitions.

Our findings have significant implications for experiments testing the effects of OA conditions on marine animals in general, illustrating the importance of testing individuals from multiple populations to assess the capacity for adaptation and response to this (and other) environmental challenges. Focusing on individuals from one site, as is typical for most studies of the effects of OA, is unlikely to capture the full potential of a species’ response, and could result in misleading conclusions. Our results are contrary to predictions that animals exposed to coastal OA conditions, or other environmental challenges, will be more robust to OA. For blue mussels, our results suggest that some populations are already robust to increased OA conditions, and they have the potential to adapt to OA as environments continue to change.

A minimum of 50 adult blue mussels (M. edulis, 30–66 mm maximum shell length) were collected from each of four populations around Long Island Sound (LIS). Populations were located in the east (Groton, G; Orient Point, O), and central parts (Milford, M, Stony Brook, S) on both the north and south shores (Fig. 1), where mussels experience very different environments. Historical carbonate chemistry for LIS is sparse, therefore, to understand the carbonate chemistry experienced by each population, monthly water samples were taken from each source location from July to January to understand seasonal and locational differences in water chemistry. In eastern LIS blue mussels experience near oceanic conditions with Ω_AR as high as 5.71, and to the west more estuarine and eutrophic environmental conditions exist (lower salinity, lower pH, and Ω_AR<1.0, Table 1).

An OA experimental system at the Northeast Fisheries Science Center in Milford CT was used for holding and conditioning adult mussels and for the larval experiments. Seawater for adults and larvae was pumped from Milford Harbor (41° 12′ 42.46″ N, 73° 3′ 7.75″ W). Adults were conditioned for spawning in a flow-through system, which is thoroughly described elsewhere (Perry et al., 2015; Meseck et al., 2021). Briefly, compressed air was scrubbed using a molecular-sieve CO₂ adsorber (Puregas) down to <10 ppm. Scrubbed air and carbon dioxide were then mixed using mass-flow controllers (Aalborg Instruments and Controls, Orangeburg, NY, USA) with each treatment bubbled into PVC columns to produce CO₂ enriched seawater. The system was designed to produce three different OA treatments (low-OA, mid-OA, and high-OA) following best experimental practices and data reporting guidelines established by EPOCA (Riebesell et al., 2011). Enriched air-CO₂ was also constantly bubbled into each corresponding OA treatment container (Table 3). Larval experiments were conducted in a static OA system with the same mass-flow controllers and air-CO₂ mixture used in the flow-through system. Filtered seawater (0.35 µm) was pre-equilibrated for 24 h prior to use, with water replacement in replicates every Monday, Wednesday, and Friday. During the experiment, the air-CO₂ mixture for each treatment was constantly bubbled into each replicate bin (five bins per treatment, three treatments) to maintain the three OA treatment levels (Table 3). From the flow-through system, water samples were taken weekly, but in the static system (larval rearing), water samples were taken before and after the water was exchanged in each replicate bin three times a week. Samples were taken in dark polypropylene bottles (500 ml) from each tank and immediately analyzed for dissolved inorganic carbon (DIC) and pH. A YSI probe (model 556, Yellowspring, OH, USA) was used to record temperature and salinity at the time of sample collection. Measurements of pH were determined at 20°C spectrophotometrically (CARY 100) with m-cresol purple following the protocol outlined in the Guide to Best Practices for Ocean CO₂ Measurements (Dickson et al., 2007). TRIS buffer obtained from the Dickson Laboratory (Scripps Institute, San Diego, CA, USA) were analyzed with each run to ensure accuracy (±0.0014). DIC measurements were analyzed using an Apollo SciTech's DIC Analyzer. Internal standards were used to calibrate the instrument and Certified Reference Material from the Dickson Laboratory was used to assure accuracy (±2.7 μmol kg⁻¹). Precision of the carbonate chemistry measurements was confirmed by an international inter-laboratory comparison exercise with the laboratory being within 0.5% of the assigned values (Bockmon and Dickson, 2015). DIC and pH were used to back-calculate the saturation state of aragonite (Ω_ar) and the pCO₂ of each treatment using CO2SYS (Pierrot et al., 2012) with the following constants: K1, K2 from Mehrbach et al. (1973) refit by Dickson and Millero (1987); K hydrogen sulfate from Dickson et al. (1990); and total Boron from Uppström (1974).

Three OA test conditions, low-OA (pH 7.9, Ω_ar 1.53, pCO2 514.50), mid-OA (pH 7.57, Ω_ar 0.77, pCO2 1167.13) and high-OA (pH 7.26, Ω_ar 0.39, pCO2 2479.74), were produced (Table 3).

Adult mussels from each source population were kept separately in screened cages in the flow-through system with natural plankton and supplemented daily with cultured microalgae (Chaetoceros sp.). Individuals from each population were divided equally among the three OA treatments and conditioned for 4 weeks. Prior to spawning, mussel shells were scrubbed to remove all attached organisms and then placed in 0.35 µm-filtered seawater. Spawning was induced by thermal shock alternating warm (up to 20°C) and cool (15°C) seawater until spawning was induced. In some cases, a male was dissected and sperm (from that population and OA treatment) was added to the water to induce spawning, and some animals were also injected with 0.5 ml of 2 M potassium chloride (KCl) to induce spawning (Strathmann, 1987). Spawning individuals were isolated in filtered sea water to keep gametes from mixing inadvertently.

The eggs from all females in a population and OA treatment combination were pooled, and then pooled sperm from males in that same population and OA treatment combination was used to fertilize eggs. Embryos for each population/OA treatment combination were placed in buckets with 15 L 0.35 µm filtered seawater (19°C, salinity 26) pre-conditioned to target OA condition, and allowed to develop for 48 h without food (larvae do not feed until the reach D stage). Two days after fertilization, larvae from each population and OA treatment level were allocated to replicate beakers (n=5).

Forty-eight hours post-fertilization larvae were reared at 300 individuals l⁻¹ in 0.35 µm-filtered sea water, fed 40,000 cells ml⁻¹ of an equal mixture of Tisochrysis lutea and Pavlova lutheri (strain MONO). We used a factorial design with larvae from four populations×three OA treatments; each treatment was replicated five times, for a total of 60 larval cultures. Each replicate culture was held in a 1-L beaker with two large-screened windows (50 µm mesh size), allowing water and microalgal food to readily pass through. One beaker from each population was held in each replicate bin (10 L, n=5) for each OA treatment. Bins were held in a circulating, temperature-controlled bath, and distributed so that no two replicates of the same treatment were adjacent to each other. When bins were cleaned and replicate beakers replaced (three times per week), survivorship and the number of individuals that metamorphosed were determined, providing relatively high-resolution data. Animals were maintained in the static system until all individuals either died or metamorphosed.

Survivorship and time-to-metamorphosis data were analyzed with accelerated failure time (AFT) models (surviminer and survival packages, R version 3.6.1). These models accommodate censored data, in which the exact timing of the event (in this case, death or metamorphosis) happening in some subset of individuals is unknown after the experiment has ended. Data from this study were right censored. For the survivorship analysis, the date of death was unknown for individuals that metamorphosed, and their time of death was after the experiment had ended. For the time to metamorphosis analysis, those that died had an unknown true time to that developmental stage. The hazard ratio was also calculated, which allowed us to compare the slopes of survivorship and development curves. We fit models using exponential, log logistic, and Weibull distributions using the survival package in R software, and then used the Akaike's Information Criterion to choose the best model. We used population (S, G, O or M) and OA treatment (high, mid, low) as fixed factors, and the replicate culture as a random factor (individuals in the same culture are not independent). Because this experiment included both fixed and a random effect, we included a “frailty” term for replicate culture in the model for both survivorship and time to metamorphosis. Frailty models in survival analysis are analogous to mixed-model analysis of variance in measuring heterogeneity at the level of the frailty term.

Funding was provided by Sea Grant R/XG-25, to D.K.P., L.M. and S.M. Numerous staff at the NOAA lab assisted with various aspects of the larval experiment. Many SBU undergraduates assisted with building larval chambers and maintaining animals, including E. Acari, M. Bradley, Y. Malakhova, K Patel, Y. Neira, and others. M. McCarty-Glenn helped with collecting animals from the field. P. Vlahos and H. Baumann provided access to the water chemistry data for Groton CT.