In situ recovery of bivalve shell characteristics after temporary exposure to elevated pCO₂

Abstract

Ocean uptake of carbon dioxide (CO₂) is causing changes in carbonate chemistry that affect calcification in marine organisms. In coastal areas, this CO₂-enriched seawater mixes with waters affected by seasonal degradation of organic material loaded externally from watersheds or produced as a response to nutrient enrichment. As a result, coastal bivalves often experience strong seasonal changes in carbonate chemistry. In some cases, these changes may resemble those experienced by aquacultured bivalves during translocation activities. We mimicked these changes by exposing juvenile hard clams (500 μm, Mercenaria mercenaria) to pCO₂ in laboratory upwellers at levels resembling those already reported for northeastern US estuaries (mean upweller pCO₂ = 773, 1274, and 1838 μatm) and then transplanting to three grow-out sites along an expected nutrient gradient in Narragansett Bay, RI (154 bags of 100 clams). Prior to the field grow-out, clams exposed to elevated pCO₂ exhibited larger shells but lower dry weight per unit volume (dw/V). However, percent increase in dw/V was highest for this group during the 27-day field grow-out, suggesting that individuals with low dw/V after the laboratory treatment accelerated accumulation of dw/V when they were transferred to the bay. Treatments also appeared to affect shell mineral structure and condition of digestive diverticula. Although treatment effects diminished during the field grow-out, clams that were pre-exposed for several weeks to high pCO₂ would likely have been temporarily vulnerable to predation or other factors that interact with shell integrity. This would be expected to reduce population recovery from short-term exposures to high pCO₂.

Introduction

It is now well known that ocean uptake of carbon dioxide is causing decreases in pH and in the concentration of carbonate ions considered essential for shell and skeletal formation by marine organisms (Orr et al. 2005). When these conditions are reproduced in laboratory environments and field enclosures, effects on growth and survival of oceanic organisms are often strong and negative (Kroeker et al. 2010). For coastal organisms, effects are also typically negative, but the patterns and drivers of carbonate chemistry in nearshore mixing zones are far more variable over space and time than in the open ocean (Cai et al. 2010; Waldbusser and Salisbury 2014; Wanninkhof et al. 2015). Prior evolutionary history and maternal or individual exposure to such variability may moderate these negative responses (e.g., Parker et al. 2012; Murray et al. 2014; Thomsen et al. 2017). However, based on current understanding of feedbacks between plankton communities and ocean biogeochemistry, surface carbonate chemistry of major ocean regions is expected to depart from present day variability envelopes within one or two decades (Keller et al. 2014; Carter et al. 2016). This is likely to occur in coastal waters of the northwest Atlantic, where influx of arctic meltwater and increased runoff are lowering alkalinity (Yamamoto-Kawai et al. 2011; Gledhill et al. 2015). These waters interact with localized coastal processes and may intensify diurnal and seasonal excursions of carbonate chemistry in estuarine shellfish habitats (e.g., Pacella et al. 2018).

Short-term coastal acidification events have reportedly reduced larval or juvenile success in commercial shellfish hatcheries and in intensively harvested wild populations (Barton et al. 2012; Green et al. 2012). These biological effects have the potential to significantly increase population-level risk to marine bivalves (Grear et al. 2020). Most studies of coastal bivalve responses to increased partial pressure of CO₂ (pCO₂) and reduced pH have focused on larval stages, with effects potentially strongest when energy allocation to calcification is limited by yolk reserves (Gobler and Talmage 2013; Waldbusser et al. 2013). Tight control of unvarying carbonate conditions has been critically important for these larval experiments (but see Clark and Gobler 2016) and may reasonably represent levels of in situ variability experienced over the short duration of these life stages (hours to a few days). For life stages that last longer (days to weeks, months, or years) and thus span a larger range of environmental conditions, reliance on short experiments and constant chemistry seems harder to defend (Keppel et al. 2016). The challenges this introduces for experimentation are important to address, since the effects of acidification on higher-level processes (e.g., population growth and persistence) cannot be established from early life stage responses alone. In M. mercenaria populations, for example, larval survival is more sensitive than adult survival to seawater acidification, but population growth is more sensitive to adult survival than to larval survival (Grear et al. 2020). This insight was made possible by formulation of a full life cycle model from field data and suggests that small impairments of upper life stage survival could cause large increases in risk. Clearly, there is a need for studies of responses in mid and upper life stages with greater emphasis on variable in situ conditions. To that end, we studied growth responses of juvenile M. mercenaria in naturally varying seawater amended with CO₂ during laboratory experiments, followed by post-treatment field grow-outs in Narragansett Bay, Rhode Island (USA).

M. mercenaria, commonly known as the hard clam or northern quahog, is a filter-feeding bivalve that occurs naturally along the east coast of the United States, ranging from the Gulf of St. Lawrence to the Gulf of Mexico (Harte 2001). Hard clams make up one of the most commercially important fisheries in the United States, with 2016 landings valued at $57M USD (National Marine Fisheries Service 2017). They are found primarily in firm mud and sand in both intertidal and subtidal substrates, where they burrow beneath the sediment surface and use their siphons for respiration and feeding on plankton filtered from the water. Adults are tolerant of a range of temperatures and salinities, but larval survival decreases at salinities below 15 ppt and juvenile and adult growth ceases in temperatures above 31°C and below 9°C (reviewed in Grizzle et al. 2001).

Laboratory studies have shown strong adverse effects of elevated pCO₂ on growth and survival of larval and juvenile hard clams (Green et al. 2004; Talmage and Gobler 2009; Talmage and Gobler 2010; Waldbusser et al. 2010; Talmage and Gobler 2011; Green et al. 2012; Gobler and Talmage 2013; Gobler et al. 2014; Miller and Waldbusser 2016), although the dominant mechanism in this and other bivalve species may be the result of pCO₂ treatment effects on Ω_A (aragonite saturation state; Waldbusser et al. 2014) or other characteristics of the carbonate system (Thomsen et al. 2015). We focused primarily on growth in our study of hard clams, but also examined effects on shell mineral structure and on histopathological markers in the digestive diverticula (DD). Responses of mineral composition to pCO₂ have been examined previously and are complex with respect to shell size and integrity (Beniash et al. 2010; Ivanina et al. 2013; Fitzer et al. 2016; Keppel et al. 2016; Leung et al. 2017). Histopathological studies of DD have a long history in bivalve ecotoxicology and provide important indicators of physiological condition and environmental stress, but appear to be quite rare in coastal acidification research except in studies of pH effects on metals toxicity (e.g., Rodriguez-Romero et al. 2014; Cao et al. 2018). Histopathological and mineralogical results in our study provided qualitative insight into observed effects of pCO₂ treatment on growth.

Methods and study sites

We conducted field grow-out experiments in 2014 and 2015 using Mercenaria mercenaria pre-conditioned at controlled pCO₂ levels in laboratory upwellers. All laboratory work was conducted at the US EPA’s Atlantic Ecology Division (AED) in Narragansett, Rhode Island, USA. All field work was conducted in the estuarine waters of Narragansett Bay where salinities are typically 28 – 32 ppt. All source water for the laboratory experiments also came from these waters through AED’s seawater system.

Laboratory upwellers

In both years of the study, juvenile Mercenaria mercenaria (~500 μm shell length) were obtained from the Aquaculture Research Corp. in Dennis, Massachusetts. These clams were placed on screens within laboratory upwellers designed to maintain constant upward seawater flow (1800 – 2000 clams per upweller). Upweller design was a modified version of those used at the Blount Shellfish hatchery at Roger Williams University (Karin Tammi, pers. comm.); each upweller consisted of a 56 cm section of 10.2 cm dia. clear acrylic pipe extending vertically through the cover of a 39 L aquarium tank (Fig. 1). Seawater was supplied through the endcap at the bottom of the cylinder, which rested on the bottom of the tank. Thus, water circulated upward through the cylinder to an outlet at the top end and then back to the aquarium tank. The clams were placed on a screen 10 cm above the bottom of the upweller. All tanks were partially immersed in an ambient seawater bath with continuous flow from Narragansett Bay. Using gentle pressurized flow through a glass pipette connected to the seawater supply, clams were stirred several times a week to prevent clumping via byssal thread attachments. This was intended to prevent large variations in clam growth within an upweller that could result from the effects of clumping on clam access to suspended food particles in the upwelled seawater. Throughout the experiments, seawater from Narragansett Bay was pumped continuously into each aquarium tank, which also had an overflow outlet, to minimize biological feedback effects on upweller chemistry (e.g., O₂ consumption and CO₂ production) and to maintain the diurnal variation in seawater chemistry that exists in the seawater supply (Grear 2016). Background levels of phytoplankton in the seawater were supplemented daily with commercial shellfish diet (Reed Mariculture Shellfish Diet 1800).

Figure 1.

End view of aquarium and upweller used for laboratory experiment. M. mercenaria juveniles (500 μm) were placed on the screen through which upward flow was continuously maintained via pumped seawater. Aquariums were immersed in a temperature-controlled seawater bath with its water level 1 cm below the aquarium overflow outlet. See text for additional details.

Upwellers in 2014 experiment.

In the smaller 2014 experiment, which primarily served to provide pre-conditioned clams for the field experiment, one upweller was assigned to each pCO₂ treatment (ambient, medium and high) without replication. Each upweller’s aquarium received the appropriate mixture of ambient outdoor air and CO₂ (research grade, 99.999% certified purity) delivered via titanium spargers. Mixing occurred in baffled gas mixing tubes that received CO₂ from mass flow controllers and ambient outdoor air from volumetric controllers. After addition of clams, air and CO₂ flow rates in the medium and high pCO₂ treatments were increased gradually over a period of five days to achieve target pCO₂ levels. Following procedures in Dickson et al. (2007), bottle samples for total alkalinity (TA) and dissolved inorganic carbon (DIC) analysis were collected in 250 ml borosilicate glass bottles on the 8^th, 13^th and 15^th day after target pCO₂ levels were reached. pCO₂ additions in the medium and high treatments were gradually adjusted downwards during the final four days prior to removal of clams from the upwellers. Juvenile M. mercenaria were grown in the upwellers from June 25 to July 21, 2014 (26 days) and then removed, digitally photographed for length measurements, subsampled for dry weights, and relocated to field experiments (see below).

Upwellers in 2015 experiment.

We used three upwellers per treatment in the 2015 experiment; seawater was supplied to all three replicates of a given treatment from a single head tank. Also, rather than using pumps within each aquarium tank, we used one pump in each head tank, splitting the flow into the inputs for each of the three replicate upweller tanks assigned to each treatment. We used repeated measurements of dissolved oxygen (DO) and temperature in each upweller to assess experiment-wide uniformity in upweller conditions. Methods of pCO₂ manipulation in these head tanks as well as monitoring and feeding were similar to those described for 2014. Gradual increase of pCO₂ in the medium and high treatments began on May 27 and continued to June 15, 2015 (16 days). Full treatments were continued 28 days to July 13, 2015. Bottle samples were collected for TA and DIC analysis on eight dates during this period using methods described above. Clams were measured to determine weights and lengths as above and then relocated to the field sites.

Field grow-outs

Pre-conditioned quahogs from the upwellers were placed in mesh bags at one site in 2014 (Potowomut) and three sites in 2015 (Potowomut, Wickford, and Dutch Harbor; Fig. 2). The Potowomut site is at the mouth of the Potowomut River within waters designated by the Narragansett Bay Estuary Program as “impacted” due to high nutrient and low DO concentrations. Nutrient loads to the bay have been declining over the past two decades due to sewage treatment plant upgrades (Narragansett Bay Estuary Program 2017), but this site is also likely affected by local runoff. The Dutch and Wickford sites are closer to the open waters of Rhode Island Sound and are both within the “lower bay” (Fig. 2), where nutrient concentrations in the main stem are lower and DO concentrations tend to be higher than in the upper bay. The three sites are similar in depth, but the Wickford site is several kilometers from the main flow of the bay’s west passage where water quality is more thoroughly monitored. We also expected lower pCO₂ and higher pH in the lower bay, based on Wallace et al. (2014).

Figure 2.

Location of study area and the three study sites, each of which consists of three subsites (dots). Three cages were installed at each subsite; each cage contained six mesh bags of 100 pre-conditioned Mercenaria mercenaria (2 bags from each pre-conditioning treatment).

2014 field experiment.

Prior to removing clams from the upwellers in 2014, nine empty steel cages (15 × 76 × 120 cm with 5 cm mesh openings) were installed at three locations (subsites) within the Potowomut site (Fig. 2). At each subsite, three of these cages were strung together in series with 5 m lengths of rope and marked with a buoy to allow retrieval. On July 18, 2014, clams from each of the six upwellers were divided into groups of 100 individuals, placed into mesh bags (mesh size = 500 μm) and returned to the tanks containing the upwellers. On July 21, they were transported in insulated coolers to the field site where two bags from each of the three upweller treatments were suspended horizontally 5 cm above the bottom of each cage (6 bags per cage; two from each treatment). This ensured that the bags were close to the sediment surface but minimized clogging of the mesh with sediment and organic matter. The allocation of all three treatments to each cage resulted in a factorial crossing of upweller treatment and subsite (3 treatments × 2 reps per cage × 3 cages per subsite = 18 bags per subsite; N = 18 × 3 subsites = 54 bags). During the grow-out period, cages were retrieved and opened 1 to 2 times per week so that the mesh bags could be gently rinsed with seawater to remove any accumulated material before being replaced to their original location. During these visits, measurements of DO, temperature, and salinity were recorded at the water surface and bottom at each subsite. Using a van Dorn sampler, water samples were collected for carbonate chemistry, chlorophyll a (Chl a), and nutrient analysis. Carbonate samples were collected in 250 ml borosilicate glass bottles with greased stoppers using methods in Dickson et al. (2007). Water for Chl a and nutrients were collected in opaque 1 L bottles and filtered immediately upon return to the lab. Filters were placed in pre-chilled acetone solution for immediate Chl a extraction (Graff and Rynearson 2011), followed by fluorometric analysis approximately 24 hours later. The filtrate was frozen in 20 ml vials and subsequently analyzed for NO₃⁻ + NO₂⁻ and NH₄⁺. All mesh bags were retrieved on August 11, 2014 (21 days after relocation to the field sites) and returned to the lab for analysis of clam length and weight.

2015 field experiment.

We repeated and expanded the experiment in 2015 to include two additional sites (Fig. 2), each with three subsites and the same 54-bag/9-cage design as used for the Potowomut site in 2014. This resulted in 3 × 54 = 162 mesh bags, each with 100 clams. These bags were set up on July 13, 2015 and relocated to the field cages the following day. All bags were retrieved on August 10, 2015 for analysis (27 days after relocation to the field sites). Cage maintenance visits and water sampling procedures were identical to those in 2014.

Carbonate chemistry

All bottle samples from the upwellers and field sites were preserved with a saturated mercuric chloride solution (0.05% final sample concentration) and refrigerated until analysis. TA was determined for upweller and field samples using a temperature-controlled titration setup with an auto-titrator (Metrohm 877) and the non-linear least squares procedure of Dickson et al. (2007), as implemented in the ‘at’ function of the seacarb R package (R Core Team 2017). DIC was determined on a Shimadzu TOC-V total organic carbon analyzer in 2014 and on an Apollo Scitech AS-C3 DIC analyzer in 2015. DIC analyses used dilutions of Certified Reference Materials (CRMs; batches 128, 137, 139 and 151, http://cdiac.ornl.gov/oceans/Dickson_CRM/batches.html) for calibration (adj. R² > 0.9999) and a secondary standard to monitor instrument drift. TA titrations used multiple determinations of the CRM in a single-point calibration. Relative errors for DIC and TA analyses were both within 0.3%, except for the 2014 DIC runs on the TOC-V, during which error exceeded 1%. During later sample runs, error for four blind samples from the same CRM provider ranged from −0.2 to 0.0% for DIC (on the AS-C3) and from 0.1 to 0.3% for TA when compared to certified values. Total pH, pCO₂, and aragonite saturation state (Ω_A) were calculated from TA, DIC, salinity, temperature, and pressure (1 atm since all samples were collected near sea level) using the seacarb R package with thermodynamic constants specified as k_f = “pf” (Perez and Fraga 1987), k₁k₂ = “l” (Lueker et al. 2000), and k_s = “d” (Dickson 1990).

Mercenaria responses

Length, weight, and volume.

Clam lengths and weights were determined two times in each of the two years during this study: immediately after the upweller phase (i.e., prior to field grow-out) and immediately after the field grow-outs in both years (2014 and 2015). Post-upweller weights were determined from sub-samples (4 subsamples in 2014; 5 in 2015) of ten individuals collected from each upweller. These subsamples were rinsed with freshwater, dried for 3 days at 60°C., and weighed. In 2014, all live clams for each mesh bag retrieved from field cages were immediately photographed. In 2015, due to the larger number of samples, clams were first preserved and then later photographed. Clams from each bag were divided into three subsets that were: 1) preserved in ethanol for length measurements; 2) preserved in 10% neutral buffered formalin (NBF) for histopathological analysis to assess morbidity and digestive cell condition; and 3) dried and weighed as above. In the ethanol set, clams with open valves were counted and removed before closed-hinge clams were counted and photographed for length measurement. All images were analyzed using customized scripts with ImageJ software (Rasband 2009). These scripts used thresholding techniques with user-supervised particle selection to count and measure length as the feret diameter (maximum distance between any two points on the perimeter) for each clam. All photographs contained scale bars for sizing and all 2015 photographs also contained gray scale reference palettes for standardizing the pixel intensity of shell edge definition in the thresholding step. Total clam volume was estimated using the relationship with clam length reported in Bricelj and Malouf (1980) for Great South Bay, NY (ln Vol = 2.9532 × ln length - 3.6657). This was used to compute dry weight per unit volume (dw/V).

Histopathological biomarkers and survival.

Each of the mesh bag subsamples that had been preserved in NBF (one subsample per bag) was assigned an accessioning number and all microscopic interpretations were completed using this number to assure impartiality to the pCO₂ treatment and field site. Samples were rinsed in 70% ethanol and initially dyed with alcoholic eosin to increase visualization during tissue processing (Peters et al. 2005). Clams were decalcified with RDO rapid decalcifier (Apex Engineering Products Corporation) and rinsed overnight in running tap water. Multiple clams per sampling were placed in labeled cassettes and processed using an Autotechnicon Mono Tissue Processor (Technicon Instruments), dehydrated with a graded series of ethanols, infiltrated with tertiary butyl alcohol into liquid 15.6°C paraffin (Paraplast Plus, McCormick Scientific), and embedded into hardened paraffin blocks. Paraffin blocks were serially sectioned at 7 μm on a Leitz microtome (Model 1512), affixed to labeled glass slides and stained with hematoxylin and eosin for microscopic analysis (Natanson et al. 2007). All clams within each serial section were examined for morbidity and any remarkable pathologies (condition of DD tissue and visceral ganglia). Despite impartiality of the observer, histopathological assessments are qualitative, so we did not plan to perform statistical analysis on these results.

Crystalline structure of shells.

In 2015, clams used for dry weight measurements after the upweller and field phases were retained for Fourier transform infrared analysis (FTIR; Silverstein and Webster 1998) of CaCO₃ crystal structure. In brief, shell material absorbs infrared energy at specific vibrational frequencies that depend on the length and geometry of the chemical bonds in the crystal structure. Clams from the upwellers were pooled by treatment whereas clams from the field were pooled by site and treatment before this analysis. From each pool, twenty randomly selected clams were measured and weighed to determine a shell/tissue ratio; the remainder was composited and ground into powder using a mortar and pestle. Known proportions of nearly pure calcite and aragonite were prepared as a reference. Powdered aliquots of 0.5 mg of each mixture and sample were measured on a Nicolet iS50 FTIR Spectrometer (Thermo Fisher Scientific, Inc) with ATR single reflection diamond ATR accessory (for the shell samples, three aliquots of each composite were analyzed). Spectra were collected in the mid-IR range from 4000 – 650 cm⁻¹ (aperture size = 100 μm); each spectrum consists of 128 scans at 8 cm⁻¹ resolution. A new background measurement was taken after every two hours of analysis to correct spectral effects of accumulating carbon dioxide and water vapor in the laboratory. All shell sample spectra were baseline-corrected before visual analysis and calculation of peak heights. There are many potential diagnostic peaks in the FTIR spectra of CaCO₃ mixtures (Loftus et al. 2015), but their interpretation often depends on the complexity of the mixture and the availability of standards and supporting alternative methods (e.g., Raman and X-ray spectroscopy). Thus, we focused on absorbances at 900 – 700 cm⁻¹ due to the presence of peaks near 856 and 713 cm⁻¹ (and their ratio) that are diagnostic of aragonite and the amorphous precursor (ACC) to crystallized CaCO₃ (Weiss et al. 2002; Beniash et al. 2010; Loftus et al. 2015).

Statistical analyses

Differences between clam lengths and weights in the 2015 upweller experiment and in the field experiments (both years) were analyzed using linear mixed effects models (lme4 package, R Core Team 2017) and likelihood ratio tests (LRTs, p = Pr > χ²) with upweller or subsite location as random effects since correlation was expected among the multiple clam measurements within each of these “subjects.” Contrasts of least squares means were performed using Tukey’s p-value adjustment, unless noted otherwise. Computation of dw/V prior to the grow-out required pooling of measurements for each upweller (n = 3 upwellers per treatment), so ANOVA was used to analyze treatment effects on dw/V. Pooling at the bag level was also required for the estimates of dw/V after the field grow-out, but a mixed effects model was used since there were 6 bags per subsite for each treatment (N = 54 and 162 bags in 2014 and 2015, respectively; see above). Analysis of clam survival used logistic regression with mixed effects. Because of the distance between field sites in 2015, we were unable to obtain simultaneous observations of water column characteristics even though we expected them to change rapidly as sunlight intensified during our early morning site visits. Thus, we included time of day as an effect in the statistical models to test whether site differences were artifacts of our sampling schedule. In the analysis of FTIR data, pooling of shells to produce powder composites meant that only one estimate per site × treatment combination was available, so we treated site as a blocking factor in the post grow-out data. Since the sites were intentionally located along a known bay gradient, this pooling introduced ‘restriction error’ (Sokal and Rohlf 1981), so we did not test for site effects on FTIR results. Pearson correlation coefficients are denoted as “r.” All p values for likelihood ratio tests (LRT) are for Chi-square tests. Degrees of freedom estimated with the Satterthwaite method are denoted as df_s.

Results

Upweller carbonate chemistry

The pCO₂ bubbling treatments in the upwellers caused clear differences in measured pCO₂, DIC, and the values of total pH. pCO₂, and Ω_A that were calculated from TA and DIC measurements (Table 1). Ambient pCO₂ levels may have been affected by respiration within the seawater plumbing system, but were not notably different from levels reported for the lower bay by Wallace et al. (2014). Mean DO in the upweller spot measurements in 2014 remained within a 0.5 mg L⁻¹ range across treatments. Mean temperatures were nearly identical in the three upwellers.

Table 1.

Characteristics of the laboratory upwellers used for pCO₂ pre-conditioning treatments prior to field grow-out of Mercenaria mercenaria.

Year	2014b			2015b
Treament	Ambient	Medium	High	Ambient	Medium	High
Salinity (PSU)	31 (0.1)a	31 (0.1)	31 (0.1)	31 (0.4)	31 (0.4)	31 (0.4)
Temperature (°C)	20.9 (0.5)	20.9 (0.5)	20.9 (0.4)	20.1 (0.7)	20.1 (0.7)	20.2 (0.6)
DO (mg l−1)	6.1 (0.5)	6.6(0.5)	6.1(0.3)	5.9(0.5)	5.7(0.7)	5.9(0.7)
TA (μmol kg−1)	2159 (48)	2222 (71)	2239 (93)	2081 (34)	2100 (38)	2093 (21)
DIC (μmol kg−1)	2017 (42)	2152 (100)	2190 (75)	1954 (27)	2038 (36)	2076 (21)
pCO2 (μatm)c	684 (176)	1222 (429)	1411(261)	677 (64)	1137 (175)	1661(134)
Total pHc	7.84 (0.11)	7.63 (0.13)	7.57 (0.09)	7.82 (0.04)	7.62 (0.06)	7.47 (0.03)
Ωaragonitec	1.8 (0.4)	1.2 (0.3)	1.1 (0.2)	1.6 (0.1)	1.1 (0.1)	(0.1)

^aAll table values are means with standard deviations.

^bThe 2014 experiment consisted of one upweller per pCO₂ level. The 2015 experiment used three upwellers per treatment.

^cpCO₂, Ω_aragonite, and total pH were calculated from measurements of DIC and TA.

Differences in carbonate chemistry between upweller treatments were larger in 2015 than in 2014 (Table 1). Mean DO and temperature in the upweller spot measurements in 2015 remained within a 0.21 mg L⁻¹ and 0.12° C range, respectively, with no differences detected between treatments in linear mixed effects analyses (df = 2, χ² = 1.16 and 0.23, p = 0.5604 and 0.8915, respectively).

Upweller effects on Mercenaria

Dry weights and lengths after the upweller phase were higher in 2015 than in 2014 (Fig. 3), possibly because of the 16 day pre-acclimation period used in 2015. Mean clam weight was highest in the medium treatment in both years. In 2015, when replication allowed statistical comparison, this difference was significant between the high and medium pCO₂ treatments (least squares mean contrasts; df = 9, t = −3.156, p = 0.0284) but not between medium and ambient treatments (df = 9, t = −1.206, p = 0.4790) or between high and ambient treatments (df = 9, t = 1.949, p = 0.1807). No clear indications of mortality (e.g., open valves) were observed in any of the upwellers in either year (see below for tissue analysis of clams with closed valves).

Figure 3.

Combined plot of 2014 (points +/− sd) and 2015 (box-whisker plots) clam dry weights, lengths, and dry weight per unit volume (dw/V; mean +/− sd for 2015) at the end of ambient, medium and high pCO₂ upweller treatments. Actual pCO₂ levels for each treatment and duration of the experiments differed between years (see Table 1). Differences in presentation are due to the difference in replication (one upweller per treatment with four weight subsamples and ~1000 lengths from each upweller in 2014; three upwellers per treatment with five weight subsamples and ~2000 lengths from each upweller in 2015).

In 2014, mean clam lengths in the three treatments were within 0.1 mm of each other (Fig. 3). In 2015, treatment effect on length was larger and significant (Fig. 3; LRT, df = 2, χ² = 13.091, p = 0.0014), with the clams from the high pCO₂ treatment being significantly larger than both the ambient (df_s = 9, t = −5.416, p = 0.0011) and the medium treatments (df_s = 8.99, t = 3.107, p = 0.0306). However, as noted above, treatment also affected dry weight. Thus, dw/V in 2015 differed significantly among treatments (F = 9.6194, Pr > F_6,8 = 0.0134) and decreased from ambient to medium to high treatments ($\overline{x}$ = 13.0, 12.0 and 7.3 mg mm⁻³, respectively). When corrected for unequal variances (Fig. 3), the difference was significant for medium vs. high treatments (df = 2.12, t = 4.88, p = 0.0351), but not for ambient vs. medium (df = 3.53, t = 0.57, p = 0.5991) or ambient vs. high treatments (df = 2.06, t = 4.04, p = 0.0535).

Effects after grow-out in Narragansett Bay

Mean seawater characteristics at the three field sites largely conformed with expected patterns, except that differences were small (Table 2). Dutch Harbor, which was the site closest to open waters of Rhode Island Sound, had the lowest temperature, chlorophyll a, and calculated pCO₂ and had the highest salinity, DIC, TA, and calculated total pH and Ω_A. Patterns were less clear for nitrate+nitrite, ammonium, phosphate, and DO. Correlation between pCO₂ and DO was strong in both years (Fig. 4) but slightly less negative in 2014 (df = 16, r = −0.69, p = 0.0015) than in 2015 (df = 34, r = −0.77, p < 0.0001). pCO₂ and DO (but not Chl a) were both correlated with time of day, which differed by 1.2 hours between Dutch and the other two sites, so time-corrected values were included in Table 2 to allow comparison between sites.

Table 2.

Environmental characteristics of the three study sites used in the grow-out of pre-conditioned Mercenaria mercenaria (mean and standard deviation).

	Potowomut		Wickford	Dutch
Year	2014	2015	2015	2015
Hour of day	9 (0.5)a	7.3 (1.0)	7.3 (2.6)	8.5 (1.2)
Salinity (PSU)	29 (0.9)	29 (0.6)	30 (0.3)	31 (0.3)
Temperature (°C)	23.5 (0.6)	23.7 (0.6)	24.3 (0.5)	22.2 (0.6)
DO (mg l−1)	6.1 (0.9)	4.6 (0.6)	5.1 (1.4)	5.9 (1.2)
TA (μmol kg−1)	1934 (40)	1953 (40)	2017 (23)	2064 (28)
DIC (μmol kg−1)	1823 (22)	1863 (40)	1898 (48)	1925 (48)
pCO2a (μatm)	766 (185)	891 (118)	798 (204)	668 (150)
Total pHb	7.76 (0.11)	7.70 (0.05)	7.76 (0.09)	7.83 (0.08)
Ωaragoniteb	1.5 (0.4)	1.3 (0.1)	1.6 (0.3)	1.7 (0.3)
Nitrate+Nitrite (μg l−1)	4 (3)	9 (5)	12 (8)	13 (8)
Ammonium (μg l−1)	13 (16)	21 (13)	27 (20)	18 (10)
Phosphate (μg l−1)	35 (12)	49 (12)	42 (12)	32 (13)
Chlorophyll a (μg l−1)	8.1 (4.3)	10.8 (3.4)	10.8 (3.0)	6.8 (2.2)
DO at 08:00 EDTc	5.4	5.0	5.5	5.7
pCO2 at 08:00 EDTc	820	855	759	693
Ωaragonite at 0800 EDTc	1.4	1.3	1.6	1.7

^aAll table values are means with standard deviations.

^bpCO₂, Ω_aragonite, and total pH were calculated from measurements of DIC and TA before averaging (see text).

^cTime-standardized DO and pCO₂ were calculated using linear interpolation to allow site comparison.

Figure 4.

pCO₂ plotted against DO during the 2014 field experiment at Potowomut (21 Jul to 11 Aug) and the 2015 field experiment at Potowomut, Wickford, and Dutch Harbor sites (14 Jul to 10 Aug). pCO₂ was calculated from measurements of DIC and TA in bottle samples.

Microscopic evaluation indicated that all closed-valve clams processed for histology were alive at the time of preservation, but their condition was affected by pCO₂ pre-treatments (see below). Based on the enumeration of closed- and open-valve clams after retrieval from the 2015 field experiment, clam survival was high (94%). There were no detectable survival effects of either the site of deployment (LRT, df = 2, χ² = 2.4892, p = 0.2881) or the pCO₂ pre-conditioning (df = 2, χ² = 0.7818, p = 0.6867).

In 2014, there were no detectable effects of pCO₂ treatment on post grow-out clam weight (df = 2, χ² = 1.7758, p = 0.4115) but a small effect on length (χ² = 11.379, p = 0.0034). Lengths were higher in the medium treatment (= 2.8 mm) than in both the ambient ($\overline{x}$ = 2.63 mm; df_s = 709, t = −3.124, p = 0.0053) and the high pCO₂ treatments ($\overline{x}$ = 2.64 mm; df_s = 709, t = −2.786, p = 0.0151). No treatment effect on dw/V was found after the field grow-out in 2014 ($\overline{x}$ = 6.34 mg mm⁻³, df = 2, χ² = 1.2093, p = 0.5463). In 2015, during which the high pCO₂ treatment was higher and the field grow-out phase was longer, there were clear effects of the pCO₂ pre-treatments on post grow-out clam weights and lengths (Fig. 5).

Figure 5.

Box-whisker plots of M. mercenaria dry weights, lengths, and dry weight per unit volume (dw/V) in the 2015 field grow-out experiment after pre-conditioning in laboratory upwellers at ambient, medium and high pCO₂. Plots represent distributions of mean clam weight and size per mesh bag (N = 162; 54 bags per site; n = 18 bags per box-whisker).

For post grow-out clam weights in 2015, the mixed effects model with all fixed effects (intercept, site, treatment, and site × treatment interaction) was significantly better when compared via likelihood ratio to reduced models with fewer fixed effects (vs. site and treatment effects only: df = 4, χ² = 11.969, p = 0.0176; vs. treatment effect only: df = 6, χ² = 22.212, p = 0.0011; vs. site effect only: df = 6, χ² = 32.513, p < 0.0001; and vs. intercept only: df = 8, χ² = 42.754, p < 0.0001). Comparison of least squares means revealed that the interaction was the result of larger treatment differences for clam weights at Potowomut and Wickford than at Dutch (Fig. 5).

For post grow-out clam lengths in 2015, the full model with the site × treatment interaction effect was not significantly different from the site + treatment model (df = 4, χ² = 5.5776, p = 0.2330), but the site + treatment model was better than reduced models with treatment (df = 2, χ² = 8.9884, p = 0.0112), site (df = 2, χ² = 25.637, p < 0.0001) or intercept (df = 4, χ² = 34.617, p < 0.0001) as the only fixed effects. Least squares means of clam lengths from the ambient pre-treatment were smaller after the field grow-out than those from the medium (df = 146, t = −3.082, p = 0.0069) and high pre-treatments (df = 151, t = −5.281, p < 0.0001) but there was no significant difference between those from medium and high pre-treatments (df = 146, t = 2.245, p = 0.0671; Fig. 5).

Effects of pCO₂ pre-treatment in 2015 on dw/V after the grow-out, using the mean clam weights with mean volumes computed from lengths, were not detected, but the model with site effect only was better than the intercept only model (Fig. 5; df = 2, χ² = 7.408, p = 0.0246). Mean dw/V was higher at Wickford ($\overline{x}$ = 15 mg mm⁻³) than Potowomut ($\overline{x}$ = 11.7 mg mm⁻³; df = 12.7, t = −2.760, p = 0.0410); no other site differences were detected (p > 0.29 for the contrasts with Dutch, for which $\overline{x}$ = 13.6 mg mm⁻³). Although Chl a concentration measured at the subsite level differed between sites (Table 2) and might indicate differences in food availability, it was not significantly correlated with subsite means for dw/V (df = 7, r = −0.19, p = 0.6304) or with the change in dw/V during the grow-out phase (df = 7, r = 0.05, p = 0.8988).

Histopathological analysis of DD in post grow-out clams was performed on 145 slides (Fig. 6). Vacuolization was noted more frequently in slides prepared from high pCO₂ pre-treatment mesh bags (20 of 49 bags; 41%) than in bags from ambient and medium treatments (6 and 16 of 48, respectively). Vacuolization was also more frequent in slides from Wickford than from the other sites. Dilation was noted with similar frequency among treatments (35–40% for all treatments) but was higher at Dutch (52%) than at Potowomut (17%) and Wickford (41%). Basophilic and eosinophilic inclusion bodies were each noted in 7 of the 145 slides. In the 22 samples where visceral ganglia were assessed, enlargement or neuron drop-out was noted in at least one slide for all treatments and all sites. Samples sizes for this result and for the inclusion bodies were too small or uneven to warrant further analysis.

Figure 6.

Examples of tissue section images of digestive diverticula from juvenile Mercenaria mercenaria after grow-out in Narragansett Bay showing normal (a), vacuolated (b), and dilated (c) condition.

FTIR spectra for the prepared mixtures of aragonite and calcite exhibited the expected double peak near wavenumber 700 cm⁻¹ in mixtures with high aragonite:calcite ratios (Fig. 7). The ratio of absorbances (A) for 713 : 700 cm⁻¹ ranged from 1.42 (aragonite:calcite ~ 0:1) to 2.04 (aragonite:calcite ~ 1:0). The highest peaks in the 875–850 cm⁻¹ region were generally associated with the lowest aragonite:calcite ratio but were not perfectly ordered. This may be due to impurities in the mixtures. Given that impurities and other calcium polymorphs are certain to exist in the shell samples, caution is needed for interpreting the peaks in this range.

Figure 7.

FTIR spectra for hard clam shells at the end of the upweller pCO₂ treatment period. The double peak near v₄ in the upper panel diminishes as aragonite:calcite ratio decreases in the known mixtures (Loftus et al. 2015). The higher peaks at v₂ in the middle panel indicate presence of amorphous CaCO₃ in the upweller clams (i.e., lower “crystallinity”; Beniash et al. 2010). There were no consistent patterns in these characteristics among treatments or sites after the field grow-out (lower panel).

After the upweller phase, mean A₇₁₃/A₇₀₀ values in shell samples were 1.23 (sd = 0.02), 1.30 (0.02), and 1.24 (0.03) for the ambient, medium and high pCO₂ treatments (n = 3 aliquots for each composited treatment; Fig 6). Mean peak heights in the 875–850 cm⁻¹ range were 0.13 (0.03), 0.21 (0.01), 0.16 (0.03) for the ambient, medium and high pCO₂ treatments, respectively. None of these mean FTIR characteristics were consistently related to mean clam weight for each treatment, although this comparison required pooling of weights into a single mean for each treatment. After the field grow-out, FTIR differences were negligible: mean A₇₁₃/A₇₀₀ values were 1.30 (0.01), 1.3 (0.01), and 1.28 (0.01) for ambient, medium and high treatments, respectively (Fig. 7). Mean peak heights in the 875–850 cm⁻¹ range after grow-out were 0.14 (0.01–0.02) for all three treatments. More than 90% of the total dry weight of clams used in the FTIR analyses consisted of shell material.

Discussion

Laboratory exposure of juvenile Mercenaria mercenaria to elevated pCO₂ caused unexpected effects on clam lengths in both 2014 and 2015, with a clear and positive effect in the larger 2015 experiment. After the ‘pre-conditioned’ clams were placed in Narragansett Bay for several weeks, some of the effects of the upweller pre-treatments persisted and, in the case of clam weight, this persistence differed between the three grow-out sites included in the 2015 experiment. Estimates of dw/V during the upweller phase were more consistent with our expectation that elevated pCO₂ would cause negative effects on bivalves. However, this negative response disappeared or became undetectable after the field grow-out. This was due to a large “recovery” of dw/V in clams from the high pCO₂ pre-treatment, which began the grow-out period with the lowest dw/V values and then underwent the largest relative weight gain and lowest volumetric growth. Since >90% of total clam dry weight consisted of shell, this suggests that these clams underwent a period of lower shell density or reduced inner shell deposition relative to clams from the ambient and medium pCO₂ treatments. Consistent with these results, the carbonate region of the FTIR spectra after the upweller treatments exhibited qualitative differences that disappeared after the field grow-out. Similar responses have been seen in other studies. For example, diminishing pCO₂ treatment differences were shown for shell area in juvenile eastern oysters grown in situ after laboratory exposure to high pCO₂ (Keppel et al. 2016). In their multigenerational study of Mytilus edulis, Thomsen et al. (2017) reported initial effects on larval size followed by recovery during the final phase of their experiments.

In the early days of the upweller phase in our study, clams were smaller than the size range that Waldbusser et al. (2010) found would allow juvenile M. mercenaria to overcome effects of aragonite undersaturation (Ω_A<1.0). However, except for the high pCO₂ treatment in 2015, mean Ω_A exceeded 1.0 in all of our pCO₂ treatments in both years and thus might explain our differing results. In addition, Waldbusser et al. (2010) pointed out that shell growth and calcification are not identical processes due to the presence of an organic matrix in shell material. Therefore, even without the plasticity in molluscan mineral architecture that has been elucidated more recently (e.g., Leung et al. 2017), shell density, rather than size, may in some cases be more directly affected by calcification. In this respect, the negative responses to pCO₂ in Waldbusser et al. (2010) and our study are more congruent. The longer duration of our pCO₂ exposure (weeks rather than hours) may also underlie differences between the two studies.

Vacuolization of the DD was more prevalent in clams from the high pCO₂ treatment after the grow-out, but the potential mechanisms for this effect are numerous. Although vacuolization is often used as a biomarker of pollutant exposure in bivalves (Neff et al. 1987; Syasina et al. 1997; Usheva et al. 2006; Rocha et al. 2016), it can also be the result of other factors, including disease (Lauckner 1983), environmental stress (Bright and Ellis 2009), and surplus lipid storage (Lowe and Clarke 1989) that may be the indirect result of contaminants, seasonal factors such as phytoplankton abundance or energy storage prior to gametogenesis (Pennec et al. 2001), or calcium ion storage (Taieb 2001). FTIR responses after the upweller phase are similarly difficult to interpret, especially given that the medium pCO₂ treatment appeared to result in a lower aragonite:calcite ratio than both the ambient and high pCO₂ treatments. In addition, attribution of spectral differences to aragonite:calcite ratios is speculative due to the potential presence of ACC on depositional shell surfaces, variable crystallinity of aragonite, and their effects on FTIR absorbances (Weiss et al. 2002). However, we can confidently state that the disappearance of qualitative treatment differences in FTIR spectra after the field grow-out occurred in parallel with diminishing differences in dw/V. These FTIR results and the tissue observations both suggest there are calcification and metabolic responses to acidification that are not well understood for juvenile bivalves.

The upweller phase of our study differed from most laboratory studies in that we used a seawater flow-through setup (rather than recirculation) and did not eliminate natural variability in carbonate chemistry before bubbling CO₂ into the incoming seawater from lower Narragansett Bay. However, carbonate variability among spot measurements was higher in the medium pCO₂ treatment (Table 1) and may have contributed to observed responses. Variability of coastal carbonate chemistry at diurnal and seasonal time scales is well known from field studies (Wootton et al. 2008; Waldbusser and Salisbury 2014; Baumann et al. 2015; Baumann and Smith 2017; Pacella et al. 2018) and has been included as a controlled variable in laboratory studies of larval bivalves (Clark and Gobler 2016) and in mechanistic modeling studies of post-larval bivalves (Miller and Waldbusser 2016), where it was found to have significant impacts on survival independent of mean conditions. We did not track diurnal variability in our study, but Grear (2016) showed pH variability in an experiment using the same laboratory space, equipment, pCO₂ control system, and seawater source as used here. For the ambient treatments, the coefficient of variation in mid-day pH measurements in that study and calculations of pH from spot measurements of DIC and TA in this study were identical (cv = 0.004).

We expected that the bayside intake and plumbing system would modify the carbonate chemistry in the seawater source, but comparison with observations in Wallace et al. (2014) indicates that the pCO₂ levels achieved in the ambient, medium, and high pCO₂ treatments were representative of mainstem bottom waters in the lower, mid and upper bay, respectively. In contrast, the treatment means spanned a much larger range (773–1838 μatm) than the field sites used in the grow-out phase (794 – 954 μatm, time-corrected), which did not differ statistically from each other. This is likely because our sites were well inshore of the main axis of the bay. In these shallow and biologically productive areas, where mean Chl a levels were above averages typically seen in temperate estuaries (Apple et al. 2008), the well-known north-south gradients of the bay were less evident, especially for nitrate and ammonium. However, the Dutch Harbor site had the highest TA, DIC, pH and DO and the lowest pCO₂ (Table 2), as expected due to its proximity to Narragansett Bay’s ocean boundary and its distance from the eutrophic waters of the upper bay. Thus, our three-site design captured some of the bay’s spatial variability and improves the generality of inferences regarding pCO₂ treatment effects on grow-out responses.

Since the clams were protected within mesh bags during the field grow-out, we were unable to test whether the period of reduced shell density or thickness (dw/V) prior to recovery from temporary exposure of juveniles to high pCO₂ could result in increased mortality from predation, as predicted by others (Gaylord et al. 2011). Sanford et al. (2014) observed increased predation on newly settled Olympia oysters (Ostrea lurida) that had been raised through the larval stage under elevated pCO₂, although they attributed the effect to prey size rather than shell density or thickness. Structural pCO₂ effects on predation vulnerability have also been observed in corals (Acropora millepora; Doropoulos et al. 2012) and in Sydney rock oysters (Saccostrea glomerata; Amaral et al. 2012). Thus, predation may be an important selective force affecting responses of shell characteristics to altered carbonate chemistry.

A change in mineral structure or an increase in inner shell calcification could result in increased clam weight without affecting clam size, and thus might underlie the rapid response of shell dw/V to field conditions. An increase in aragonite:calcite ratio, for example, would result in higher dw/V. Although we did not observe a clear or linear correspondence between dw/V and the FTIR response, the qualitative differences in carbonate region of the FTIR spectra for the three pCO₂ upweller treatments appeared to diminish during the field grow-out. Leung et al. (2017) argued that mineralogical plasticity allows calcifying organisms to acclimate to acidifying conditions based on their finding that the length of southern periwinkles (Austrocochlea constricta) increased more rapidly in elevated pCO₂ treatments, leading to larger shell volume per unit mass. Likewise, Ivanina et al. (2013) observed negative effects of pCO₂ on shell hardness in adult Eastern oysters (Crassostrea virginica) and hard clams when combined with elevated temperature. Such structural changes relate to ‘crystallographic disorder,’ which Fitzer et al. (2014; 2018) reported as a response to elevated pCO₂ in adult blue mussels (Mytilus edulis) and Sydney rock oysters (Saccostrea glomerate). They also observed increases in calcite thickness in M. edulis at the highest pCO₂ treatment (1000 μatm for 6 months). In juvenile Eastern oysters, Beniash et al. (2010) observed responses to elevated pCO₂ in calcite lath thickness, reductions in shell hardness, and a reduction in shell weight but not length. This suggests a reduction in shell density or thickness similar to the one we observed in hard clams, but which contrasts with the maintenance of density rather than size observed in Olympia oysters by Sanford et al. (2014). They suggested that shell thickness may be favored over shell size as a predation deterrent, but we note that there may also be predation-driven selective pressure for rapid growth in size at the expense of shell thickness. In any case, the diversity of reported responses is consistent with results from Busch and McElhany (2017), who found that species sensitivity within mineralogical and phylogenetic groupings of Puget Sound calcifiers was variable. It seems possible that the dominant predation modes affecting each bivalve prey species would be a useful grouping factor to consider for predicting which shell characteristics will be conserved, if any, in response to acidification.

The ability to distinguish between plasticity, acclimation, and adaptation of bivalves to changing carbonate chemistry is essential to predicting the likelihood of total or commercial extinction under elevated pCO₂ conditions. In our study, hard clams responded quickly when relocated to the bay. We found no evidence that pre-exposure to high pCO₂ enhanced final hard clam condition at the end of the field grow-out phase, although none of the field sites had pCO₂ levels as high as the high pCO₂ treatment. Mean aragonite saturation state (Ω_a) was lower in the field than in the upweller for only one of the nine treatment × site groups (Potowomut clams from the ambient pCO₂ upwellers; Tables 1 and 2). Clams from the high pCO₂ treatment experienced the largest improvement in Ω_a when moved to the bay (+0.5, +0.8 and +0.9 for the three sites) and thus would have been the least acclimated to the bay’s Ω_a conditions, yet these clams underwent the largest percent increase in dw/V (78% vs 9% and 11% for the ambient and medium treatments). Thus, rapid compensatory response to bay conditions rather than pre-acclimation appears sufficient to explain the fast recovery of dw/V in pre-exposed juvenile hard clams. Selection can also be easily ruled out because mortality was negligible in the upweller phase. In their multi-generational study of M. edulis, Thomsen et al. (2017) pointed out that temporally varying selective pressures such as those experienced by early life stage bivalves should reduce allele fixation or favor plastic phenotypic responses such as those in our study. Thomsen et al. (2017) was referring primarily to transgenerational effects on yolk phase larvae, but in nearshore areas, seasonal overlap of the early post-larval stage with favorable conditions is also expected to be variable given effects of precipitation, temperature, and wind events on carbonate parameters and mixing. This is not testable with our data, but the methods we used could conceivably be combined with multi-generational approaches and adapted to that purpose.

Much interest in coastal acidification has centered around the occurrence of high pCO₂ (or low aragonite saturation state) during sensitive periods of bivalve life cycles (Gobler and Talmage 2013; Waldbusser and Salisbury 2014; Grear et al. 2020), as has been observed at our study site (Wallace et al. 2014). Similarly, shellfish aquaculture often contends with short-term variation in local conditions, especially in hatcheries (Barton et al. 2012). Our upweller treatments followed by grow-out in less extreme ambient conditions were partially intended to simulate the short-term exposures that can occur in unmanaged populations as well as in aquaculture operations. Our results suggest that deficiencies in shell characteristics induced by a few weeks of exposure of juvenile hard clams to pCO₂ levels in the 1500–2000 μatm range, which already exist seasonally in northeastern US estuaries (Wallace et al. 2014), may diminish within several weeks of growth in a favorable field setting, but potential effects on predation vulnerability or compensatory physiological processes associated with calcification demands remain unknown.

Scientific Significance Statement.

Most studies of coastal acidification impacts on bivalves have been performed on larval stages with constant exposure levels in laboratory settings. This study examined effects of temporary exposure on post-larval juvenile hard clams by first pre-exposing them to elevated carbon dioxide in the laboratory and then translocating them to an estuarine setting. Clams showing shell deficiencies immediately after the exposure recovered those characteristics during the month-long field grow-outs. Despite this ability to respond rapidly to improved conditions, seasonal coastal acidification events may cause brief windows of vulnerability to predation during early life stages.