Physiological responses of three species of unionid mussels to intermittent exposure to elevated carbon dioxide

Zones of elevated carbon dioxide (CO2) have been proposed as barriers to fish movements, but impacts of CO2 on non-target sensitive organisms are unknown. In the present study, freshwater mussels were exposed to intermittent elevations in CO2 and showed species-specific mechanisms to regulate acid–base disturbances, along with a stress response.


Introduction
Environmental levels of carbon dioxide (CO 2 ) that are commonly found in freshwater ecosystems have the potential to act as both continuous and intermittent stressors for aquatic organisms. Over the past several decades, levels of CO 2 in the atmosphere have been increasing as a result of the anthropogenic burning of fossil fuels, which has led to a concomitant increase in the partial pressure of CO 2 gas (pCO 2 ) in marine ecosystems (Shirayama and Thornton, 2005). Unlike marine systems, there is no consensus regarding how pCO 2 will change in freshwater as a result of climate change (Hasler et al., 2016). In freshwater, pCO 2 can vary across and within watersheds (Butman and Raymond, 2011), as well as episodically and on seasonal and diel cycles within water bodies (Maberly, 1996). In a review of 7000 global rivers and streams, the average median value for pCO 2 was~3100 µatm (Raymond et al., 2013) 1 679 ± 543 to 35 617 ± 46 757 µatm, with means in the USA ranging from 679 ± 543 to 9475 ± 993 µatm (Cole and Caraco, 2001). In addition to these natural sources of elevated pCO 2 , recent work has shown that zones of elevated CO 2 can act as non-physical fish barriers, thereby providing a management tool to prevent the movement and spread of invasive fish species (Kates et al., 2012;Noatch and Suski, 2012). Although a specific method for the use of CO 2 barriers to deter fish movement has not yet been defined, one potential application is the intermittent addition of CO 2 into a navigational lock or approach channel at vulnerable times (i.e. when lock doors are open; United States Army Corps of Engineers, 2014a), resulting in downstream pulses of CO 2 -rich water. Thus, downstream fluctuations in CO 2 might occur, making CO 2 a potential intermittent stressor for freshwater organisms.
A taxonomic group of freshwater organisms that may be particularly at risk to CO 2 stressors are freshwater mussels (Order Unionoida). Mussels serve many important ecological functions, influence many ecosystem processes (Vaughn and Hakenkamp, 2001) and are often used as indicators of ecosystem health (Williams et al., 1993). Although North American freshwater ecosystems contain the highest diversity of freshwater mussels in the world (Williams et al., 1993;Bogan, 2008), more than half (71%) are listed as endangered, threatened or of special concern, largely as a result of anthropogenic stressors, such at habitat alteration and degradation (Williams et al., 1993;Ricciardi et al., 1998). Additionally, while mussels are generally considered a homogeneous group of sessile animals, there are four main tribes of mussels in North America (Quadrulini, Lampsilini, Pleurobemini and Amblemini) that all vary in morphology, physiology and reproductive strategies and may thus respond differently to environmental stressors.
At present, there is a paucity of research on the effects of elevated pCO 2 on freshwater invertebrates, particularly unionid mussels. Hannan et al. (2016a, b) found that mussels experience acid-base regulation in response to short-and long-term exposures to elevated pCO 2 , and a stress response to long-term exposure to elevated pCO 2 . Previous studies on marine bivalves indicate that elevated pCO 2 causes internal acidosis (Michaelidis et al., 2005;Bibby et al., 2008) that is often buffered by increasing HCO 3 − in the fluids (Pörtner et al., 2004). Both marine (Michaelidis et al., 2005) and freshwater mussels (Hannan et al., 2016a, b) can increase haemolymph HCO 3 − by using CaCO 3 released from the shell as a result of decreased pH and elevated CO 2 (i.e. increases both haemolymph HCO 3 − and Ca 2+ ) or by reducing the activity of the Cl − -HCO 3 − exchanger to retain HCO 3 − at the cost of Cl − uptake (Byrne and Dietz, 1997;Hannan et al. 2016a, b). Another strategy to buffer acidosis is to alter the activity of Na + -H + exchangers to increase removal of H + ions, thus also increasing Na + uptake (Byrne and Dietz, 1997;Lannig et al., 2010, Hannan et al. 2016b). Exposure to a chronic elevation in pCO 2 also appears to initiate the general stress response in mussels, because a decrease in haemolymph Mg 2+ and an increase in haemolymph glucose have been observed in unionid mussels (Hannan et al. 2016a, b). More importantly, previous studies (i.e. studies described above) that have quantified CO 2 stressors in mussels have used a continuous application of CO 2 rather than one that was intermittent as might be expected downstream of a CO 2 barrier, and differences may exist between the continuous application of a stressor relative to one applied intermittently (exacerbation, attenuation or no change; Reinert et al., 2002).
Based on this background, the goal of the present study was to quantify the physiological impacts of intermittent exposures to elevated pCO 2 on three species of freshwater mussels each belonging to a different tribe, Pyganodon grandis (tribe Anodontini), Amblema plicata (tribe Amblemini) and Lampsilis cardium (tribe Lampsilini). To accomplish this goal, over a 28 day period the mussels were exposed to either control pCO 2 or intermittent increases in pCO 2 and then sampled for a suite of physiological parameters related to acid-base status and physiological stress. The results of this study help to clarify further how different exposures to elevated pCO 2 affect the acidbase and stress responses of various freshwater mussel species in habitats where pCO 2 fluctuates.

Mussel collection and husbandry
Plain pocketbook (L. cardium) and threeridge mussels (A. plicata) were collected by benthic grab from the Mississippi River, Cordova, IL, USA, in July 2015. Giant floater mussels (P. grandis) were collected by benthic grab from a barrow pit near Champaign, IL, USA, in August 2015. Mussels were taken to the Aquatic Research Facility at the University of Illinois, Champaign-Urbana, IL, USA in coolers (travel time <3 h for L. cardium and A. plicata and <1 h for P. grandis). Upon arrival at the Aquatic Research Facility, all mussels were cleaned of epibionts and tagged for individual identification with a permanent marker (Neves and Moyer, 1988). Once tagged, mussels were placed in three tubs (1136 litres) supplied with water from a 0.04 ha natural, earthen-bottom pond, where they remained for at least 1 week to recover from transport stressors and to acclimate to laboratory conditions (Dietz, 1974;Horohov et al., 1992;Dietz et al., 1994). All tubs were equipped with a Teco 500 aquarium chiller (TECO-US, Aquarium Specialty, Columbia, SC, USA) and a low-pressure air blower (Sweetwater, SL24H Pentair, Apopka, FL, USA) to maintain aeration. Fifty per cent water changes using pond water were performed weekly to maintain water quality. Mussels were fed a commercial shellfish diet of the following consituents: Nannochloropsis sp. 1-2 µm and a mixed diet of Isochrysis, Pavlova, Thalassiosira and Tertraselmis spp. across all holding tanks with a portable meter (YSI 550A, Yellow Springs Instruments, Irvine, CA, USA) and averaged 22°C (21.7 ± 0.1°C, mean ± SEM) and 7.50 mg l −1 (7.60 ± 0.06 mg l −1 ). Water pH was measured using a handheld meter (WTW pH 3310 meter, Germany) that was calibrated regularly, and averaged 8.55 ± 0.01 throughout the acclimation period. Dissolved CO 2 and total alkalinity (TA) concentrations were measured using digital titration kits and averaged 4.86 ± 0.04 mg l −1 and 1093.0 ± 27.0 µmol kg −1 , respectively (Hach Company, Loveland, CO, USA; Titrator model 16,900 catalogue no. 2272700 and catalogue no. 2271900 for CO 2 and TA, respectively).

Fluctuating CO 2 exposure
To define the impacts of fluctuating CO 2 on mussel physiology, mussels (L. cardium, A. plicata and P. grandis; n = 28) were separated into two recirculating treatment systems (92 litres), each with nine 5 litre tanks (adapted from Hohn and Petrie-Hanson, 2007). Systems were maintained as stated above with the exception that one system received a CO 2 treatment. In the CO 2 treatment system, pCO 2 was turned on every 1.5 h, and increased from ambient (~1000 µatm, 1355 ± 119 µatm; pH = 7.85 ± 0.02) to~55 000 µatm (56 492 ± 1342 µatm; pH = 6.62 ± 0.03) by bubbling CO 2 gas into the water through an air stone (see Supplementary material, Fig. S1). Elevated pCO 2 was held constant at~55 000 µatm for 0.5 h, for a total of 12 fluctuations per day. Thus, animals were held at elevated pCO 2 levels for 0.5 h and returned to control levels during the 1.5 h recovery period and then raised back up to elevated conditions for 0.5 h, repeatedly during the course of the experiment. A level of 55 000 µatm was targeted because this level has previously been defined as being a potential target CO 2 level that could deter the movement of fishes (Donaldson et al., 2016) and will possibly be the target level of a CO 2 barrier. Twelve fluctuations per day represents the historical lock usage of Brandon Road Lock (41.5054°N, 88.0996°W), a possible site for deployment of a CO 2 barrier within the Des Plaines River, IL, USA (United States Army Corps of Engineers, 2014aEngineers, , 2015. The target pCO 2 was maintained with a pH controller (PINPOINT ® , American Marine Inc., CT, USA) that automatically bubbled CO 2 into the tank system through an air stone should the pH rise above a target level during exposure (Reynaud et al., 2003;Riebesell et al., 2010). The level of CO 2 was then returned tõ 1000 µatm by bubbling in air though an air stone to off-gas excess CO 2 . An identical recirculating system was used as a control, and mussels in this control system were treated in the same way as animals receiving CO 2 , except that infused CO 2 gas was replaced with compressed air such that mussels were held continuously at ambient~1000 µatm (876 ± 108 µatm; pH = 8.13 ± 0.02) pCO 2 . A digital timer (DT620 Heavy Duty Digital Timer, Intermatic Inc Spring Grove, IL, USA) was used to control additions of CO 2 and air. A modified infrared probe was used to measure pCO 2 (Vaisala GMP220 and GMT221, Vantaa, Finland; Johnson et al., 2010), along with a CO 2 titration kit to determine the concentration of CO 2 (Hach Company, catalogue no. 2272700, Loveland, CO, USA). Before and after the 12.00 h exposure, temperature (21.7 ± 0.1°C) and DO (7.60 ± 0.07 mg l −1 ) were measured as stated above, and the temperature, pH (see above) and TA (2566.3 ± 252.9 µmol kg −1 ) were entered into CO2calc to verify pCO 2 (Robbins et al., 2010).
Individual mussels were non-lethally and repeatedly sampled for haemolymph on day 1, 4, 7, 14, 21 or 28 of exposure to fluctuating pCO 2 or control conditions. Mussels were sampled during the 1.5 h period when CO 2 was at ambient levels, not during the 0.5 h when CO 2 levels were elevated. Prior to starting this study, it was not known whether sampling mussels immediately prior to the increase in CO 2 or immediately after the period of increased CO 2 would be optimal to define the impacts of CO 2 on physiological parameters. Therefore, mussels were sample during both intervals, and n = 7 animals were sampled immediately prior to the increase in CO 2 , whereas a second n = 7 animals were sampled immediately following the increase in pCO 2 , once pCO 2 returned to control values. All samples were collected around the 12.00 h CO 2 exposure to standardize any potential for diel variation in physiological parameters.
Haemolymph (0.5 ml for L. cardium and A. plicata; and 0.25 ml for P. grandis) was extracted from the anterior adductor muscle with a 1 ml syringe and 26 gauge needle (Gustafson et al., 2005) and then centrifuged at 12 000g for 2 min. After centrifugation, the supernatant was removed, flash frozen in liquid nitrogen and stored at −80°C until processing. Mussels were sampled for haemolymph only once per sampling day, and were randomly sampled before or after the CO 2 exposure on each sampling day over the 28 day period. On day 28 of exposure, mussels were sampled for haemolymph as stated above and then lethally sampled. Lethal mussel sampling included measurements for length, width, depth of the whole mussel using digital callipers (traceable digital carbon fiber calipers, Fisher Scientific, Pittsburg, PA, USA), and weight of the whole mussel (tissue + shell) was collected to the nearest 0.01 g using a balance (HL-300WP, A&D, Ann Arbor, MI, USA). Soft tissue dry weight (in milligrams) was determined by taking mussel soft tissues and drying them at 99°C for 24 h before weighing (Widdows et al., 2002). If possible, sex was determined for L. siliquoidia and P. grandis using both their external sexual dimorphism and by examination of the gills for glochidia (Trdan, 1981).
The dry weight and length of individuals within each species was not different between control and fluctuating CO 2 treatment groups (Student's unpaired t-test, P > 0.05; Table 1). Additionally, mortalities were limited over the exposure period, but occurred for two and five P. grandis from the control and fluctuating pCO 2 treatments, respectively, and for two A. plicata and one L. cardium exposed to the fluctuating pCO 2 treatment.

Statistical analyses
The effects of CO 2 exposure on haemolymph ion levels and glucose concentrations were quantified using a two-way analysis of variance (ANOVA), with pCO 2 (fluctuating or control), sampling day and their interaction (pCO 2 × sampling day) entered into each model as fixed effects. Individual mussel identification number (ID), time point (i.e. sampling before or after pCO 2 ), length, dry weight and sex (if applicable) were initially included in the models as cofactors to quantify their potential influence on response variables, but were removed because they had no significant effect on model outputs (Engqvist, 2005;Zuur et al., 2009). If at least one of the main effects in the ANOVA model was significant, or if the interaction term was significant, a Tukey-Kramer honestly significant difference (HSD) post hoc test was applied to separate means (Rohlf and Sokal, 1995). Finally, a separate Student's unpaired t-test was run on each species to quantify differences in dry weight and length across different pCO 2 treatments.
For all statistical analyses, analysis of fitted residuals using a quantile-quantile plot (Anscombe and Tukey, 1963) was used to assess normality, while a Hartley's F max test (Hartley, 1950), combined with visual inspection of the distribution of fitted residuals, was used to assess homogeneity of variances. If either normality or homogeneity of variance assumptions were violated (Siegel and Castellan, 1988), data were rank transformed Table 3: Results of two-way ANOVA examining the impact of fluctuating exposure to elevated pCO 2 on Amblema plicata exposed to one of two different pCO 2 treatments [~1000 µatm (ambient); intermittent at~55 000 µatm] for 28 days and then re-analysed within the same parametric model described above, and the assumptions of both normality and equal variances were confirmed (Conover and Iman, 1981;Iman et al., 1984;Potvin and Roff, 1993). All data are presented as means ± SEM where appropriate, all tests were performed using R (version 3.2.2), and differences were considered significant if α was <0.05. For all variables and species, there was no effect of sampling before vs. after CO 2 application (i.e. time point) for either treatment (control or fluctuating), so data from mussels sampled before and after CO 2 application were combined.

Results
There was a significant interaction between treatment and day for all three species of mussels for haemolymph HCO 3 − (Tables 2-4). At 14 days of exposure to fluctuating pCO 2 , P. grandis (treatment × day, F = 13.0, P < 0.001; Fig. 1A) and A. plicata (treatment × day, F = 8.61, P < 0.001; Fig. 1D) had approximately a 2-fold increase in haemolymph HCO 3 − relative to mussels held at ambient pCO 2 , and these concentrations remained significantly elevated for the duration of the exposure period. For L. cardium, haemolymph HCO 3 − was significantly elevated beginning at 4 days of exposure compared with control mussels and throughout the rest of the exposure period (treatment × day, F = 0.52, P < 0.001; Fig. 1G).
A significant interaction of treatment and day was also found for all three species of mussels with respect to haemolymph Ca 2+ concentrations (Tables 2-4). Pyganodon grandis had a significant elevation in haemolymph Ca 2+ concentrations beginning at 14 days of exposure to fluctuating levels of CO 2 relative to control mussels; however, it should be noted that

6
control mussels also experienced a decrease in haemolymph Ca 2+ concentrationsat 28 days compared with 1 day of treatment (treatment × day, F = 8.42, P < 0.001; Fig. 1B). For A. plicata, haemolymph Ca 2+ in mussels exposed to fluctuating levels of CO 2 was significantly elevated compared with control mussels at 7 days of exposure (treatment × day, F = 4.14, P = 0.003; Fig. 1E). A similar increase in haemolymph Ca 2+ in L. cardium occurred in response to fluctuating pCO 2 , where Figure 1: Concentrations of HCO 3 − , Ca 2+ and Cl − in the haemolymph of Pyganodon grandis (n = 9-14; A-C), Amblema plicata (n = 12-14; D-F) and Lampsilis cardium mussels (n = 13-14; G-I) exposed to two treatments of pCO 2 ,~1000 µatm (control) or intermittent increase at 55 000 µatm for 1, 7, 14, 21 or 28 days. Data are presented as means + SEM. *Groups that were significantly different from the control treatment within a time point (two-way ANOVA; see Tables 2-4). For (I), there was no significant interaction between pCO 2 treatment and sampling day; a bar above the treatments of a day represents a significant effect of time (two-way ANOVA).
For P. grandis and A. plicata (Tables 2 and 3), there was a significant interaction between treatment and day for haemolymph Cl − concentrations, but only a significant effect of day for L. cardium (Table 4). Haemolymph Cl − was lower in P. grandis exposed to fluctuating pCO 2 at 7 days of exposure relative to control mussels; however, it is important to note that this might have been attributable to a significant increase in haemolymph Cl − concentrations in control mussels at 7 days compared with control mussels on day 1 of treatment (treatment × day, F = 5.66, P < 0.001; Fig. 1C). In A. plicata, although there was a significant interaction of treatment and day, haemolymph Cl − was not significantly different in mussels exposed to fluctuating levels of CO 2 and control mussels at any point throughout the exposure (treatment × day, F = 3.63, P = 0.008; Fig. 1F). In L. cardium, no significant effect of CO 2 treatment was detected, and haemolymph Cl − increased overall at 28 days of treatment (day, F = 5.73, P < 0.001; Fig. 1I).
With respect to haemolymph Na + , a significant interaction between treatment and day was detected for all three species of mussels (Tables 2-4). Pyganodon grandis exposed to fluctuating pCO 2 had a significant elevation in haemolymph Na + at 28 days of exposure compared with control mussels (treatment × day, F = 3.15, P = 0.017; Fig. 2A). Haemolymph Na + for both A. plicata (treatment × day, F = 12.85, P < 0.001; Fig. 2B) and L. cardium (treatment × day, F = 3.05, P = 0.0194; Fig. 2C) exposed to the fluctuating pCO 2 were significantly elevated compared with control mussels beginning at 4 days and throughout the duration of the exposure period.
For haemolymph Mg 2+ , a significant interaction between treatment and day was also found for all three species of mussels (Tables 2-4). Haemolymph Mg 2+ was significantly reduced at 14 and 21 days of exposure to fluctuating pCO 2 compared with control mussels for P. grandis (treatment × day, P. grandis, F = 28.33, P < 0.001; Fig. 3A) and A. plicata (treatment × day, F = 15.68, P< 0.001; Fig. 3B), but these concentrations were no longer different from control mussels at 28 days of exposure. Likewise, haemolymph Mg 2+ in L. cardium exposed to fluctuating CO 2 levels was significantly decreased compared with control mussels on 7 and 14 days but returned to control values after 21 days of exposure (treatment × day, F = 7.31, P < 0.001; Fig. 3C).
For haemolymph glucose, there was no significant interaction of treatment and day for any species of mussel (Tables 2-4). Haemolymph glucose concentrations of P. grandis and L. cardium were unaffected by pCO 2 exposure (treatment, P > 0.05; Fig. 4A and C). Haemolymph glucose of A. plicata was significantly affected by fluctuating pCO 2 treatment, but not sampling day, and was elevated throughout the exposure period compared with control mussels (treatment, F = 8.75, P = 0.004; Fig. 4B). Figure 2: Concentrations of Na + in the haemolymph of Pyganodon grandis (n = 9-14; A), Amblema plicata (n = 12-14; B) and Lampsilis cardium mussels (n = 13-14; C) exposed to two treatments of pCO 2 , 1000 µatm (control) or intermittent increase at~55 000 µatm for 1, 7, 14, 21 or 28 days. Data are presented as means + SEM. *Groups that were significantly different from the control treatment within a time point (two-way ANOVA; see Tables 2-4).

Discussion
Following exposure to fluctuating elevated pCO 2 , all three mussel species demonstrated physiological changes indicative of disturbance in acid-base regulation. Exposure to high CO 2 often causes the acidification of internal fluids in aquatic animals (Pörtner et al., 2004), and one strategy for animals to buffer this internal acidosis is to increase HCO 3 − concentrations (Pörtner et al., 2004;Pörtner, 2008). Marine mussels are thought to increase haemolymph HCO 3 − by using CaCO 3 released from the shell (Michaelidis et al., 2005;Bibby et al., 2008). In the present study, all three mussel species may have used this buffering strategy during intermittent pCO 2 exposure, as both haemolymph HCO 3 − and Ca 2+ were elevated. Haemolymph HCO 3 − can also be increased by reducing activity of Cl − -HCO 3 − exchangers, thus increasing the retention of HCO 3 − in the haemolymph but at the cost of Cl − uptake (Byrne and Dietz, 1997). A reduction in haemolymph Cl − , which appears to be a short-term response to elevated pCO 2 (Hannan et al., 2016a, b), was observed only in P. grandis; however, this difference may have been due to rising Cl − concentrations in the haemolymph of control mussels rather than a decrease in Cl − of CO 2 -treated mussels. A third strategy that mussels use to buffer acidosis is to mediate activity of Na + -H + exchangers to increase excretion of H + ions, thus also increasing Na + uptake (Byrne and Dietz, 1997;Lannig et al., 2010;Hannan et al. 2016b). Increases in haemolymph Na + were observed for all species of mussels exposed to fluctuating pCO 2 , but the timing of the elevation in haemolymph Na + was species specific, as haemolymph Na + concentrations for A. plicata and L. cardium mussels increased after 4 days and for P. grandis after 28 days of exposure. Together, the results of the present study suggest that the three species of mussels used similar mechanisms to deal with acidosis to marine mussels (i.e. manipulating haemolymph HCO 3 − and H + concentrations); however, speciesspecific differences in these responses occurred.
In addition to an acid-base disturbance, the results of our study indicate that the stress response was also activated. An indicator of stress in freshwater mussels is declining Mg 2+ of the haemolymph, which has been associated with stressors such as elevated temperature (Fritts et al., 2015a), exposure to heavy metals (Hemelraad et al., 1990) and chronic exposures to elevated pCO 2 (Hannan et al., 2016a, b). Haemolymph Mg 2+ concentrations decreased by~2-fold in all mussel species exposed to the fluctuating pCO 2 treatment, but returned to control values after 28 days of exposure. In contrast, Hannan et al. (2016a) did not observe a return of Mg 2+ to control values in Fusconaia flava exposed to~20 000 µatm pCO 2 for 32 days. In addition, Lampsilis siliquoidea but not A. plicata exposed to either 20 000 or 55 000 µatm pCO 2 showed a decrease in haemolymph Mg 2+ during 28 days of exposure, and these values returned to baseline once the CO 2 stressor was removed (Hannan et al. 2016b). Although the pCO 2 and the species of mussels were not the same in our study and those of Hannan et al. (2016a, b), these data suggest that Figure 3: Concentrations of Mg 2+ in the haemolymph of Pyganodon grandis (n = 9-14; A), Amblema plicata (n = 12-14; B) and Lampsilis cardium mussels (n = 13-14; C) exposed to two treatments of pCO 2 , 1000 µatm (control) or intermittent increase at~55 000 µatm for 1, 7, 14, 21 or 28 days. Data are presented as means + SEM. *Groups that were significantly different from the control treatment within a time point (two-way ANOVA; see Tables 2-4

9
fluctuating exposures to elevated pCO 2 have a different effect on the Mg 2+ response of unionid mussels compared with a chronic exposure. Haemolymph glucose concentrations, another indicator of stress in freshwater mussels (Patterson et al., 1999;Fritts et al., 2015a), were elevated only in A. plicata. Increasing glucose in response to stress comes at a cost to nonvital functions, such as growth, reproduction and movement (Patterson et al., 1999;Fritts et al., 2015a). Although the interaction of pCO 2 and sampling time was not significant in the model, the elevation in glucose in A. plicata appeared to return to control levels following 28 days of exposure to fluctuating pCO 2 , suggesting that A. plicata recovers in terms of this stress marker by the end of the exposure period. A similar increase in haemolymph glucose was also observed for A. plicata exposed to a chronic elevation in pCO 2 at 55 000 µatm over a 28 day period (Hannan et al. 2016b), suggesting that fluctuating and long-term exposure to pCO 2 may have similar effects on haemolymph glucose in this species. Taken together, changes in haemolymph Mg 2+ and glucose concentrations suggest that all three species of mussel experienced physiological stress during exposure to fluctuating pCO 2 ; however, desensitization, acclimation or recovery might have occurred over extended exposure to the intermittent CO 2 stressor.
Physiological changes, such as acid-base and stress responses, experienced by animals following a stressor are energetically challenging, and long-term upregulation or maintenance of these responses can lead to less energy availability for nonvital functions, such as growth and reproduction (Wendelaar Bonga, 1997). Following exposure to intermittent or repeated stressors, animals may respond to subsequent exposures in different ways (i.e. exacerbation, attenuation or no change; Reinert et al., 2002). Our results suggest that the duration and CO 2 concentration used in the present study did not permit recovery between pulses of high pCO 2 , evidenced by the fact that mussels sampled before and after the CO 2 exposure were not statistically different from each other. Additionally, the responses of mussels to intermittent pCO 2 exposure (i.e. elevations of Ca 2+ and Na + and reduction in Mg 2+ ) were similar to those observed in unionid mussels exposed to a chronically elevated pCO 2 (Hannan et al., 2016a, b), suggesting that mussels react to the intermittent and chronic CO 2 exposures in a similar way. However, differences in the responses of these variables during intermittent (present study) and chronic exposures (Hannan et al., 2016a, b) did arise during the later stages of the 28 and 32 days exposure period, respectively. For instance, as mentioned above, the concentration of Mg 2+ returned to control values by the end of the intermittent CO 2 exposure, whereas in previous studies using either chronic exposure to elevated CO 2 (Hannan et al., 2016a, b) or elevated temperature (Fritts et al., 2015b), Mg 2+ remained reduced throughout the exposure period. In addition, haemolymph Ca 2+ (P. grandis and L. cardium) and Na + (all three mussel species) remained elevated for the intermittent exposure to 55 000 µatm pCO 2 , whereas these ions returned to control values by 32 days of chronic exposure to~20 000 µatm pCO 2 for F. flava (Hannan et al., 2016a). This sustained increase in Figure 4: Concentrations of glucose in the haemolymph of Pyganodon grandis (n = 9-14; A), Amblema plicata (n = 12-14; B) and Lampsilis cardium mussels (n = 13-14; C) exposed to two treatments of pCO 2 ,~1000 µatm (control) or intermittent increase at 55 000 µatm for 1, 7, 14, 21 or 28 days. Data are presented as means + SEM. For (B), there was no significant interaction between pCO 2 treatment and sampling day; *significant effect of pCO 2 treatment between mussels exposed to fluctuating~55 000 µatm and those exposed to~1000 µatm (two-way ANOVA; see Table 3

10
haemolymph Ca 2+ and Na + , as well as the difference in the dynamics of the haemolymph Mg 2+ response in at least two of the mussels species, may suggest that mussels respond differently to intermittent and chronic CO 2 exposure. These responses also do not exclude the possibility that the differences might be species specific or driven by the difference in pCO 2 used in these two studies. The present study suggests that exposure to intermittent elevations in pCO 2 do result in acid-base disturbances and stress responses in unionid mussels that are both attenuated (e.g. Mg 2+ ) and exacerbated (Ca 2+ and Na + ).
Species-specific responses observed in the present study might have resulted from a combination of differences in the physiology and behaviour of the three mussel species examined. Haemolymph Ca 2+ was elevated in both P. grandis and L. cardium for more than half of the treatment period, whereas Ca 2+ concentrations were elevated only on day 7 of exposure in A. plicata, suggesting that these species may rely differently on shell CaCO 3 stores. In addition, a decrease in haemolymph Cl − was observed only in P. grandis and did not occur in either L. cardium or A. plicata. These differences in the studied unionid mussels suggest that they may use different strategies to retain HCO 3 − for acid-base regulation. Finally, similar elevations in haemolymph Na + throughout nearly the entire pCO 2 exposure period were observed in L. cardium and A. plicata, whereas haemolymph Na + in P. grandis was elevated only at 28 days of exposure. This difference in haemolymph Na + concentrations in response to pCO 2 exposure suggests that L. cardium and A. plicata may rely on increased regulation of the Na + -H + exchanger to buffer acidosis, a mechanism that may be less important for P. grandis until CO 2 exposure is extended. In terms of measures of the stress response, a similar transient decrease in Mg 2+ was observed across all species; however, haemolymph glucose was elevated only in A. plicata. Taken together, similar responses to intermittent elevation in pCO 2 were observed across the three species examined, and the species differences that arose highlight the importance of considering multiple species when testing an organism's reaction to a stressor.
Results obtained in our study increase the understanding of responses of freshwater unionid mussels to fluctuating exposures of elevated pCO 2 , as modelled after a CO 2 barrier to invasive fish movement. There is evidence that, like marine mussels, if freshwater unionid mussels are exposed to elevated pCO 2 at either chronically high levels (Hannan et al., 2016a, b) or intermittent elevations (pres study), they will experience acid-base disturbances. If unionid mussels were to be exposed to intermittent elevations in pCO 2 for an extended period of time, populations might be negatively affected owing to the increased energy demand of acid-base regulation and stress responses that may come at the expense of growth and reproduction. Additionally, resident mussel species may be affected differently, as evidenced by the observed species-specific responses to elevated pCO 2 , which may arise because of differences in their behaviour and physiology. It is also important to consider that fluctuating elevations in pCO 2 may have similar but potentially also differential impacts compared with chronic exposures of elevated pCO 2 and that, generally, exposure time and duration between applications of a stressor are important aspects to consider for study design. Taken together, the results of our study suggest that the duration and manner of pCO 2 exposure (i.e. chronic vs. intermittent), as well as the species characteristics of resident unionid mussels that may be impacted, are all important factors to consider when designing, implementing and assessing the potential impacts of a CO 2 barrier.

Supplementary material
Supplementary material is available at Conservation Physiology online.