Carbon Dioxide Toxicity to Zebra Mussels (Dreissena polymorpha) is Dependent on Water Chemistry

Carbon dioxide (CO2) is gaining interest as a tool to combat aquatic invasive species, including zebra mussels (Dreissena polymorpha). However, the effects of water chemistry on CO2 efficacy are not well described. We conducted five trials in which we exposed adult zebra mussels to a range of CO2 in water with adjusted total hardness and specific conductance. We compared dose–responses and found differences in lethal concentration to 50% of organisms (LC50) estimates ranging from 108.3 to 179.3 mg/L CO2 and lethal concentration to 90% of organisms (LC90) estimates ranging from 163.7 to 216.6 mg/L CO2. We modeled LC50 and LC90 estimates with measured water chemistry variables from the trials. We found sodium (Na+) concentration to have the strongest correlation to changes in the LC50 and specific conductance to have the strongest correlation to changes in the LC90. Our results identify water chemistry as an important factor in considering efficacious CO2 concentrations for zebra mussel control. Additional research into the physiological responses of zebra mussels exposed to CO2 may be warranted to further explain mode of action and reported selectivity. Further study could likely develop a robust and relevant model to refine CO2 applications for a wider range of water chemistries. Environ Toxicol Chem 2024;43:1312–1319. Published 2024. This article is a U.S. Government work and is in the public domain in the USA. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.


INTRODUCTION
Zebra mussels (Dreissena polymorpha Pallas 1771), native to the Eastern European Ponto-Caspian region, have spread across much of Europe and North America (Gallardo et al., 2013).In North America, zebra mussels have established in many waterbodies across the eastern United States including the Laurentian Great Lakes and have more recently been detected west of the Continental Divide (US Bureau of Reclamation, 2015; US Geological Survey [USGS], 2022).Suitable habitat and successful establishment of zebra mussel populations in a waterbody are governed by thermal regime, salinity, calcium, pH, and productivity (Cohen, 2005;McMahon, 1996;Naddafi et al., 2011).The potential distribution of zebra mussels is likely shifting due to climate change and may pose a risk to areas previously thought uninhabitable for zebra mussels (Drake & Bossenbroek, 2004;Hellmann et al., 2008;Petsch et al., 2021).
Zebra mussels are tolerant to a range of physiochemical properties.They can tolerate water temperatures up to 30 °C for extended periods but thrive in cooler conditions (Churchill et al., 2017;Spidle et al., 1995).Adult zebra mussels are reported to have a minimum pH limitation of pH 6.5, but populations are more prevalent in alkaline conditions, with a pH > 7.4 identified as the threshold for veliger survival (McMahon, 1996).Calcium is an important mineral for shell development and has been described in correlation to population density (Jones & Ricciardi, 2005;McMahon, 1996).Calcium concentrations below 8 mg/L limit dreissenid mussel growth and survival (Jones & Ricciardi, 2005).Zebra mussels are more salinetolerant relative to many other freshwater bivalves, surviving in waters as high as 12‰ and thriving in brackish waters (Mackie & Schloesser, 1996).However, zebra mussels are less tolerant of ion-depleted waters, requiring higher concentrations of Ca 2+ , Mg 2+ , Na + , and K + than unionid mussels (Dietz et al., 1994).
The physiochemical parameters that delineate suitable zebra mussel habitat may also influence mussel susceptibility to a toxicant.Perhaps the most understood are the parameters included in the biotic ligand model (BLM) to predict dissolved metal toxicity.The BLM predicts changes in metal toxicity by quantifying the influence of dissolved organic carbon, dissolved ions, and pH on metal binding at target sites of aquatic organisms (Naddy et al., 2002;Paquin et al., 2000;Santore et al., 2001).The BLM was originally developed for ecotoxicology but may also be applied for predicting the efficacy of metalbased pesticides in aquatic habitats (e.g., copper).Examination of water chemistry parameters and their influence on the toxicity of other control tools, such as carbon dioxide (CO 2 ), is sparse.Whereas the relationship between alkalinity and dissolved CO 2 concentration has been well described in aquatic environments, zebra mussel tolerance of hypercapnia may depend on additional water chemistry parameters, such as dissolved ions.
Carbon dioxide is known to decrease the pH in aqueous solutions depending on the alkalinity of the system (Hargreaves & Brunson, 1996;Wurts & Durborow, 1992).Elevated CO 2 results in elevated carbonic acid concentrations through perturbation of the bicarbonate equilibrium.The decreased pH from an elevated CO 2 concentration produces acidosis in organisms and disrupts cellular processes by altering intracellular pH and ion regulation (Pynnonen, 1990).Physiological processes to overcome intracellular pH shifts are highly reliant on osmoregulation and ion exchange.Many freshwater species rely heavily on HCO 3 -for intracellular buffering, in comparison with marine species, which rely more on the exchange of acid equivalents with ions such as Na + (i.e., Na + /H + exchange; Byrne & Dietz, 1997;Horohov et al., 1992).The intracellular buffering mechanism of zebra mussels, along with other dreissenids and corbiculids (all invasive to North America), is thought to be conserved from their more recent lineage from marine species, unlike the more diverged unionid mussels (Byrne & Dietz, 1997;Horohov et al., 1992;McMahon, 1996).Carbon dioxide is currently labeled and used for the control of invasive carp (US Environmental Protection Agency [USEPA], 2019).Carbon dioxide has also been demonstrated to be an effective control tool for other aquatic invasive species including American bullfrog (Lithobates catesbeianus Shaw 1802; Abbey-Lambertz et al., 2014), crayfish (Abdelrahman et al., 2021;Fredricks et al., 2020), and sea lamprey (Petromyzon marinus Linnaeus 1758; Dennis et al., 2016).Furthermore, CO 2 has demonstrated a selectivity for zebra mussels compared with native unionid mussels.Waller & Bartsch (2018) described high zebra mussel mortality rates (80-100%) at ≤477.5 mg/L, which inflicted only approximately 1% mortality in a juvenile unionid species (Lamsilis siliquoidea Barnes 1823).In a later study (Waller et al., 2020), two other unionid species (Lampsilis cardium Rafinesque 1820 and Leptodea fragilis Rafinesque 1820) were further examined alongside zebra mussels, and a similar CO 2 selectivity was reported.Longer exposures of up to 28 days to CO 2 has been observed to induce mortality in unionid species (L.siliquoidea and Lampsilis higginsii Lea 1857), resulting in 28-day median lethal concentrations (LC50s) of ≥71,000 µatm, with some less-than-lethal responses observed at lower concentrations (Waller et al., 2019).
Investigating CO 2 toxicity to adult zebra mussels in differing water chemistries is essential if CO 2 is to be considered an effective control tool across the infested range.To address this new science, we exposed adult zebra mussels to CO 2 in reconstituted water with adjusted total hardness and specific conductance to identify the parameter(s) that influence CO 2 efficacy.

Test organisms
We collected adult zebra mussels (Wisconsin Department of Natural Resources permit #SCP-FM-2021-038) from navigation pool 8 of the Mississippi River (La Crosse, WI, USA) in October 2021.Mussels were collected from docks at public boat launches by severing their byssus with a scraping tool and were transferred to the laboratory in coolers containing temperatureacclimated well water.Mussels were held in a flow-through system at 12°C and fed continuously with blended algal feed (Shellfish Diet 1800 and Nanno 3600; Reed Mariculture) to maintain approximately 2 mg/L as dry mass.Dissolved oxygen was monitored in the mussel holding tank and was at an acceptable level throughout the study (ASTM International, 2013).
Mussels in the study ranged from 10 to 20 mm in shell length.Before each acclimation period, we assessed the viability of mussels by manually testing adductor muscle resistance to opening.Mussels that did not resist opening or failed to close were discarded.Suitable mussels (n ≥ 20) were indiscriminately placed in plastic mesh bags (15 × 15 × 2.5 cm) containing a polycarbonate tile (12 × 12 cm).Bagged mussels were acclimated to the trial temperature and reconstituted water for 7 days in a recirculating raceway.The temperature was adjusted by ≤3 °C/day to the exposure temperature (17 °C), and water chemistry was adjusted by replacing approximately 20 L of tank water (~20%) daily with water adjusted to the trial parameters.Feeding was stopped 72 h before exposure, to minimize loading the exposure tanks with waste (i.e., feces and psuedofeces).Immediately before each trial, we assessed suitability for testing based on attachment status.Mussels that had failed to attach to the bag or tile were deemed unsuitable for testing and were discarded.The selection process resulted in 19 or more mussels in each bag.
After each trial, mussel bags were consolidated into the control tanks, and the inflow was increased to maintain dissolved oxygen concentration.Feeding was resumed and consisted of approximately 0.3 mL of concentrated feed/tank daily.Mortality was assessed at 96 h post exposure and was defined as failure to resist opening when gentle, opposing pressure was applied to the valves.For each trial, the maximum anterior-posterior shell length of 40 indiscriminately selected mussels from the controls was measured with a digital caliper (CD6-ASX; Mitutoyo).

Study design
We used a continuous flow, proportional diluter system to create a range of CO 2 concentrations with a mean dilution factor of 0.73 (Cupp et al., 2020).An automatic manifold system (220 series; Harris) delivered a steady flow of food-grade CO 2 (Airgas USA) to a fine-pore air stone in the first diluter chamber at 0.7 to 0.9 L/min and 25 PSI.The diluter system delivered 17 °C water at 98 mL/min (range: 44-140 mL/min) to each 10-L tank (n = 24; Pentair).Tanks were arranged in a randomized, four-block design with each block containing five CO 2 treatments and a control.
Source water differed for each of the five trials (Table 1).We adjusted water total hardness and alkalinity by reconstituting deionized water in six bulk tanks (4000 L each) following recipes for moderately hard and very hard water (USEPA, 2002).We modified the USEPA guidelines by substituting NaCl for KCl at nominal concentrations above 4 mg/L KCl to reduce the likelihood of KCl toxicity confounding the effects of CO 2 (Fisher et al., 1991).Conductivity was adjusted by metering a concentrated stock of NaCl (50 g/L) into the headbox of the diluter system, modifying the well water to the desired specific conductance.Source water samples were analyzed for total hardness, alkalinity, specific conductance, and major cations.Total hardness and alkalinity were determined by titrimetric methods (Methods 2340 C and 2320 B, repectively; American Public Health Association [APHA] et al., 2017).Specific conductance was measured with a calibrated multimeter and probe (HQ40d; Hach).Calcium (Ca 2+ ), Mg 2+ , Na + , and K + concentrations were determined with inductively coupled plasma-optical emission spectroscopy (5110 VDV; Agilent).
During the 96-h exposures, we used a titrimetric method to determine CO 2 concentrations daily in each tank (Method 4500-CO 2 ; APHA et al., 2017).Additionally, we measured dissolved oxygen, temperature, and specific conductance daily in each tank.We measured pH three times daily in each tank as a proxy to monitor CO 2 concentrations (see the Supporting Information).Dissolved oxygen, pH, and specific conductance were measured with a calibrated multimeter and appropriate probes (HQ40d; Hach).Temperature was measured with a verified digital thermometer (Mk4; ThermaWorks).Total ammonia as nitrogen (TAN) was measured spectrophotometrically in the highest concentration at the termination of CO 2 exposure and monitored daily during the postexposure period for each trial (DR3900 and TNT 830 test vials; Hach).We used KCl as a reference toxicant to assess changes in mussel resilience due to holding conditions (see the Supporting Information).

Data analysis
All summaries and statistical analyses were conducted in R Ver. 4.1.1(R Core Team, 2021).Dose-response curves were generated with the drc package with the proportion of dead mussels as the dependent response variable and CO 2 concentration as the independent dose variable (Ritz et al., 2015).Observations were weighted within the model based on the number of mussels in each exposure tank.The experimental units were the exposure tanks (n = 24/trial).We used Akaike's Information Criterion to select the most parsimonious models.We considered the log-logistic, log-normal, and Weibull twoparameter models from the self-starter functions in the drc package (LL.2,LN.2, and W2.2, respectively).An analysis of variance test compared a single, aggregate model and individual trial models to determine differences in dose-response curves.The aggregate model was used as the null hypothesis (i.e., there were no differences in dose-response between trials), and the data were modeled together in a single curve.The independent trial model represented the alternative hypothesis (i.e., there were differences in dose-response between trials), and the data were modeled by trial with coefficients estimated for each trial independently.Dose-response models were further compared with pairwise comparisons to identify similar and dissimilar trials.We compared the confidence intervals of the LC50 and LC90 estimates (i.e., concentrations resulting in 50% and 90% mortality, respectively) of each trial to examine shifts in CO 2 toxicity.The LC50 is most commonly used for comparisons of efficacy, but we included the LC90 because it more closely resembles the objectives of typical zebra mussel treatments (i.e., 50% mortality is not considered an effective treatment).We used linear regression to examine trends between the predicted LC50 and LC90 estimates and water chemistry parameters.The LC estimates from each trial were used as the dependent variables, and independent variables included alkalinity, total hardness, specific conductance, and cation concentrations.We included the days since collection as a candidate variable to examine possible effects on toxicity of holding period.Experimental units for the linear-regression models were the trials (n = 5).All statistical analyses were interpreted at α = 0.05.

RESULTS
Temperatures during the trials deviated by <1 °C from the 17 °C target temperature (range: 16.3-17.7°C).Mean temperature during the postexposure period was 17.2 °C (range: 16.4-18.6°C).There was a negative correlation between CO 2 and dissolved oxygen (R 2 = 0.8492, F (1, 598) = 3367, p < 0.0001), but dissolved oxygen concentrations were >6 mg/L (~63% saturation) in all trials and postexposure periods.Carbon dioxide and pH were highly correlated, but the relationship differed slightly between trials based on alkalinity (see the Supporting Information).For all trials, TAN was ≤0.372 mg/L TAN in the high concentrations at the end of the exposure and ≤0.471 mg/L TAN during the 96-h postexposure period, well below critical thresholds for freshwater mussels (USEPA, 2013).Mortality in the controls did not exceed 4% in any trial and was acceptable per ASTM International methods (ASTM International, 2013).The two-parameter Weibull function was selected as the best fit and most parsimonious representation of the data (Table 2 and Supporting Information).Modeling independently by trial was significantly different and more parsimonious than modeling all the trials in a single, aggregate model (Table 3 and Supporting Information).All trials generated unique dose-response curves (p < 0.0001; Figure 1).The trials were more divergent for the LC50 estimates and more similar for the LC90 estimates.The well water trial had the lowest LC50 estimate, 108 mg/L CO 2 , and the moderately hard water trial had the lowest LC90, 156 mg/L CO 2 (Table 4).In contrast, the high conductivity trial was the least lethal at both endpoints, with an LC50 estimate of 179 mg/L CO 2 and an LC90 estimate of 217 mg/L CO 2 .Estimates of the LC90 were most similar between the very hard water and high conductivity trials (214 and 217 mg/L CO 2 , respectively), and both were less lethal than the other trials by >25 mg/L CO 2 .
Linear regression of the LC50 and Na + concentration returned the only statistically significant correlation (Table 5 and Figure 2).The model predicts an LC50 increase of 15.7 mg/L CO 2 for every additional 1 mmol/L Na + .Specific conductance was the only parameter to have a statistically significant correlation with LC90 estimates (Table 6 and Figure 2).Likewise, the LC90 model indicates that an increase of 0.07 mg/L CO 2 is required to maintain efficacy for every unit increase in specific conductance (µS/cm at 25 °C).Although not statistically significant, specific conductance shared a similar relationship with LC50, with a slope coefficient of 0.09.
A statistical difference was observed between the initial and final reference toxicant exposures (see the Supporting Information).However, the length of time the mussels were held before exposure (i.e., days since collection) did not explain the variation seen in LC50 or LC90 estimates between the CO 2 trials.

DISCUSSION
We observed differences in zebra mussel mortality after CO 2 exposure in the various water chemistries used for our trials.Our results indicate that Na + concentration is the best predictor of LC50 shifts, and specific conductance is the best predictor of LC90 shifts.Zebra mussels were more susceptible to CO 2 in waters more deplete of Na + when considering the LC50.The estimation of the LC50 provides a means of comparable lethality to other species but is of little value to control efforts when complete mortality is often the goal.Examination of the LC90 provides a more realistic estimation of the concentrations that would be used in control efforts.Our results indicate that specific conductance best predicts differences in LC90 estimates.Given the collinearity between the specific conductance and Na + and the use of conductance in determining salinity (Method 2520 B;APHA et al., 2017), the application of specific conductance as the means to adjust treatment dosage is more practical.Analysis of Na + would provide a more accurate measure of LC50 shifts, but the analysis fails to provide an immediate measure, which would require time, accessibility to instrumentation, and more funding.Comparatively, measuring specific conductance requires a calibrated meter and probe that measures the variable in real time and is far less expensive than the instrumentation required to measure Na + .One caveat to the selection of Na + in our modeling procedure is our use of NaCl disproportionally to KCl to avoid K + toxicity when adjusting specific conductance, which resulted in a wider range of Na + values being considered and may be responsible for Na + being selected over other ions that contribute to conductance.In addition, there are other ions not examined in the present study that also contribute to conductance, which may need to be considered in future research.
The greater tolerance of zebra mussels to CO 2 in water with higher Na + concentrations may be explained by osmoregulation and acid-base balance mechanisms.Zebra mussels are likely utilizing Na + /H + exchange mechanism to regulate acidosis induced by CO 2 exposure rather than using HCO 3 buffering to the extent seen in unionid mussels (Byrne & Dietz, 1997;Horohov et al., 1992).Utilization of Na + for buffering CO 2 -induced acidosis would explain the correlation between Na + and efficacy observed in our study.Although we did not capture intracellular/extracellular ion concentrations in the present study, it is a consideration for future research and may further explain the mode of action, influences of water  chemistry on efficacy, and the selectivity reported between zebra mussels and unionid species exposed to CO 2 (Waller & Bartsch, 2018).A similar trend was described for KCl toxicity being dependent on Na + concentration (Moffitt et al., 2016).Unlike CO 2 , however, the role of Na + in KCl toxicity is dependent on Na + /K + balance instead of buffering capacity (Fisher et al., 1991;Moffitt et al., 2016).The differences observed in KCl toxicity may be a result of acclimation after collection.There is evidence that hemolymph Na + is drastically reduced from stresses during collection and transport, taking approximately 18 days to return to precatch concentrations (Martem'yanov, 2000).The first reference toxicant trial was initiated 13 days after collection, which may have been while the mussels were still depleted of Na + , and the final exposure was initiated 89 days after collection when mussels may have had restored ionic balance.An altered ion balance could easily have skewed the comparison and would support the initial exposure results as being more toxic since KCl toxicity is dependent on Na + /K + balance (Fisher et al., 1991;Moffitt et al., 2016).However, we failed to collect the necessary data to investigate mussel ion balance and how it may have affected dose-response in the present study, but it highlights a potential confounding factor for laboratory testing with mussels.Recent suggestions from an Invasive Mussel Collaborative work group are to hold mussels for a minimum of 7 days to observe mortality and then hold them for an additional 48 h after they have acclimated to test conditions (Waller et al., 2023).Potassium chloride and CO 2 toxicity are likely not directly comparable because of differing modes of action.However, we did include the number of days since collection as a predictive variable in our model selection process, and it was statistically insignificant at predicting shifts in CO 2 efficacy.Unfortunately, resources did not permit an additional trial in well water at the end of our study to investigate potential temporal shifts in resiliency to CO 2 within the same source water, as was done with KCl.
A limiting factor of our investigation was the use of mussels from a single population acclimated to the desired water FIGURE 1: Dose-response curves for zebra mussel mortality when exposed to CO 2 for 96 h in different water chemistries.Water chemistries were adjusted for total hardness and specific conductance (Table 1).chemistry instead of testing mussels from different sources with different water chemistries.Moffitt et al. (2016) described similar concerns about acclimating mussels to a water chemistry different from their source and the implications of disrupting ion regulation, which could possibly confound the results.Our results in the well water trial are similar to those reported by Waller and Bartsch (2018).They reported ≥80% mortality at 96 h at concentrations as low as 157.3 mg/L CO 2 , and our models estimated an LC80 of 160 mg/L CO 2 , with both experiments utilizing the same water and mussel sources, albeit our test temperature was approximately 5 °C warmer.Regardless, we think our results highlight the implications for the application of CO 2 as a control tool and stress the importance of characterizing water chemistry when determining efficacious treatment concentrations.For example, we observed 90% mortality in our well water trial at CO 2 concentrations that would only have induced 50% mortality in "moderately conductive" water.This finding implies that no universal dose can produce a desired effect and that treatments could be more efficacious if CO 2 concentrations were to be adjusted based on Na + concentration and, which is perhaps more practical for field application, on specific conductance with considerations for the alkalinity of the water being treated.Further research could result in a robust and relevant model for adjusting CO 2 dosages for zebra mussel control applications.
Supporting Information-The Supporting Information is available on the Wiley Online Library at https://doi.org/10.1002/etc.5864.
Acknowledgements-Our research was funded with US Geological Survey (USGS) appropriated Ecosystem Mission Area apropriated funding.The authors thank C. Kirkeeng for analyzing dissolved cations, J. Smerud for assistance with constructing and calibrating the diluter system, R. Nelson for laboratory assistance, J. Schueller and two anonymous USGS reviewers for their constructive reviews of early drafts, and the anonymous journal reviewers for their time and critique of this article.
Conflict of Interest-The authors have no conflicts of interests to declare.
Disclaimer-Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.
Author Contributions Statement-Matthew T. Barbour: Methodology; Investigation; Formal analysis; Visualization; (A) ( B) FIGURE 2: Linear relationships of the estimated median lethal concentration (LC50; A) and 90% lethal concentration (LC90; B) for zebra mussels exposed to CO 2 with respective independent variables, Na + concentration, and specific conductance.Shaded area represents the 95% confidence interval for each model.

TABLE 1 :
Mean water chemistry parameters for source waters used in trials (Barbour et al., 2023)previous study utilizing the same water source(Barbour et al., 2023).

TABLE 2 :
Two-parameter, dose-response model selection for adult zebra mussel mortality from 96-h exposures to carbon dioxide in different water chemistries

TABLE 3 :
Analysis of variance comparison of single and trial-independent dose-response curves for zebra mussel mortality from 96-h exposure to dissolved carbon dioxide Water chemistry and CO 2 efficacy-EnvironmentalToxicology and Chemistry, 2024;43:1312-1319 AIC = Akaike's Information Criterion.

TABLE 4 :
Lethal concentration (LC) estimates (mg/L CO 2 ) for zebra mussels exposed to CO 2 for 96 h in various water chemistries and the highest concentration of CO 2 observed for each trial CI = confidence interval.

TABLE 5 :
Linear model summaries of examined variables to the ob- AIC = Akaike's Information Criterion.

TABLE 6 :
Linear model summaries of examined variables to the observed 90% lethal concentration for 96-h exposures of zebra mussels to CO 2 AIC = Akaike's Information Criterion.