Introduction

Calcification underlies the ecological success of numerous marine species that precipitate calcium carbonate (CaCO3) shells or skeletons. Yet, the process of calcification is sensitive to shifts in the dissolved carbonate chemistry of seawater. For example, ocean acidification (OA), which arises as anthropogenic carbon dioxide (CO2) enters the oceans and decreases seawater pH1, generally impairs the formation and integrity of CaCO3 structures2. Responses are not uniform, however, and species can exhibit increasing, decreasing, or non-monotonic changes in calcification rate as dissolved CO2 concentrations rise3. This variation creates challenges in dissecting the environmental drivers of shell or skeleton formation in marine taxa, a difficulty exacerbated by strong correlations among the multiple parameters that characterize the carbonate system of seawater (Fig. 1).

Fig. 1: Examples of the interconnectedness of the components of the carbonate system.
figure 1

Many components of the carbonate system are intertwined. As a result, the combination of total alkalinity (TA) and dissolved inorganic carbon (DIC), perhaps the two most common master variables of the carbonate system, can be used to calculate (a) pH, (b) carbon dioxide (CO2) concentration, (c) bicarbonate ion concentration (HCO3−), and (d) carbonate ion concentration (CO32-). Values shown here were calculated using a salinity of 34 and a temperature of 13.5 oC. Concentrations of CO2, HCO3−, and CO32− are expressed as µmol kg−1 and pH is quantified on the free scale. Note that within the range of TA and DIC shown, all calculated variables appear nearly correlated with each other, apart from [HCO3−] (c). Because many researchers work with CO2 in units of µatm, we note that [CO2] of 100 and 10 µmol kg−1 correspond to 2550 and 255 µ atm, respectively, at this salinity and temperature.

Full size image

Contemporary models linking calcification to environmental conditions focus on parameters within the seawater carbonate system that are thought to operate as dominant drivers4,5. For example, calcium carbonate saturation state (Ω) is often used as a metric of calcification potential. This parameter is proportional to the product of carbonate (CO32−) and calcium (Ca2+) ion concentrations, with [CO32−] operating as the more important determinant in the open ocean given that offshore calcium concentrations vary minimally. As Ω increases, especially when it exceeds a numerical value of one, shell or skeleton production is increasingly favored. The Ω of seawater also depends on the specific mineralogical form of CaCO3 (e.g., aragonite or calcite6). The relevance of Ω has been advanced in part from energetic considerations surrounding rapid precipitation kinetics7,8,9,10,11 though dissolution thermodynamics of CaCO3 are also described by Ω. Conditions of Ω <1 indicate a tendency for CaCO3 to dissolve6, though dissolution when Ω > 1 is possible for biologically formed material12.

Because Ω aligns most closely with dissolution processes but still appears to predict calcification rates, an alternate conceptualization of CaCO3 production places less emphasis on CO32− and instead forefronts HCO3− as the probable inorganic carbon source for shell or skeleton formation. This perspective notes that HCO3− ion transporters have been identified while ones for CO32− have not4,13,14, and highlights that such transporters could bolster inorganic carbon concentrations within the calcifying space. The above alternate conceptualization also emphasizes the cost of removing protons (H+) from the calcifying location, when bicarbonate operates as the source of inorganic carbon15,16. The tradeoffs between facilitation of calcification by bicarbonate and opposition by [H+] have additionally been expressed as a numerical quotient, computed as the ratio of HCO3− to proton concentrations ([HCO3−]/[H+] = substrate inhibitor ratio, or SIR)15,17,18. Using this ratio conjoins individual effects of [HCO3−] and [H+] into a single composite parameter, an easily missed departure from the conceptual origins of SIR which invoked separate lines of action for these quantities (see progression of ideas in Refs. 15,17,19.). As we detail below, important implications emerge from melding – versus considering separately – these two components of the carbonate system.

Due to the nature of the equilibrium constants of the seawater carbonate system, Ω and SIR co-vary tightly5,17. Such covariation means that calcification patterns explained by Ω can also be explained by SIR, and vice versa, neglecting for the moment whether either are ideal predictors of calcification. The numerical alignment of Ω and SIR holds most strongly when variation in salinity, temperature, and pressure is limited, and when shifts in carbonate chemistry derive primarily from changes in [CO2]. However, in the face of human-induced environmental alterations17,20, and particularly in shallow coastal zones where freshwater inputs can be important, increases in seawater [CO2] occur alongside deviations in salinity and temperature5,17,21. Shifts in these properties alter relationships among carbonate system parameters (Fig. 2a), including disrupting the otherwise linear covariance5,17,21 of Ω and SIR (Fig. 2b). Further complexities derive from freshwater and terrestrial alkalinity input to the coastal oceans alongside biologically driven chemical fluxes characteristic of shoreline habitats that can change carbonate system trajectories expected in global change scenarios22,23,24,25.

Fig. 2: Relationships among carbonate system parameters are complex in coastal oceans.
figure 2

Coastal seawater chemistry can depart considerably from broader, ambient conditions characteristic of waters offshore. a Example ranges of carbonate ion (CO32-) concentration (proportional to aragonite saturation state) and bicarbonate (HCO3−) concentrations observed within coastal zones and estuaries in natural systems, with contours for pH quantified on the free scale, and CO2 concentration in µmol kg−1. Data from tidepools in Bodega Bay, CA63, Mission-Aransas Estuary, TX79, Everglades National Park, FL80, and weekly averages from National Oceanic and Atmospheric Administration (NOAA) buoys at Cape Elizabeth81, Ocean Station Papa82, and Kuroshio Extension Observatory83 processed as described previously5. Dashed lines indicate constant-pH (free scale) contours at a temperature of 13.5 oC and a salinity of 34, and arrows demonstrate the direction of action for various processes common in coastal systems. Ambient conditions are represented as 2350 µmol kg−1 total alkalinity and 2140 µmol kg−1 dissolved inorganic carbon (DIC), typical of open ocean waters equilibrated to a pCO2 of approximately 400 µatm. For these example processes, effects of respiration and ocean acidification (Resp) increase dissolved CO2 from 400 to 1200 µatm (similar to end of century predictions for CO2) while photosynthesis (Photo) decreases dissolved CO2 to 280 µatm. The magnitude of this process arrow is consistent with a return to preindustrial CO2 levels and assumes negligible mineralization or production of organic matter. Effects of calcification (Calc) and dissolution (Diss) here reduce or raise, respectively, alkalinity by 300 µmol kg−1. The effect of rain (Rain) dilutes seawater with water containing 0 µmol kg−1 alkalinity to a final salinity of 27 while high alkalinity river runoff (River) shows the effect of mixing seawater with fresh water containing 4000 µmol kg−1 alkalinity and 2000 µmol kg−1 [Ca2+] to a final salinity of 27 and allowing pCO2 to return to 400 µatm. b Comparison of SIR and Ω for the same field data, demonstrating how the usual tight correlation between these two parameters can break down in natural waters5.

Full size image

No single model of calcification, whether based on Ω, SIR, or another parameter or combination of parameters, is likely to apply to all marine groups17,26. It is well-known that shell formation depends not only on the form of inorganic carbon used (i.e., CO32− or HCO3− or CO2) and an organism’s ability to regulate calcifying fluid pH, but also on the manner in which carbon is provisioned (e.g., direct uptake from seawater, diffusion of respiratory CO2, ion pumping)4,26,27. Independent influences of inorganic carbon uptake28,29,30,31 including saturation of the uptake pathway32,33, inhibition by increased seawater [H+] (i.e., decreased pH)34, independent actions of both of these processes20,35, and separate influences of [CO32−]36,37 have also been proposed to be drivers of calcification in a variety of autotrophic taxa. These complexities highlight the limitations of using a single parameter like either Ω or SIR to predict calcification across multiple species or across broader groups like autotrophs and heterotrophs.

Given the breadth of unknowns concerning drivers of calcification, we undertook an experiment in which we decoupled multiple components of the seawater carbonate system to isolate their effects on shell formation. We focused on adults of the California mussel, Mytilus californianus, a dominant space holder and key ecological community member along the west coast of North America38. This species is a natural choice for exploring mechanisms of shell formation because it encounters large natural variability in seawater carbonate chemistry, associated with spatiotemporal differences in ocean productivity and upwelling, as well as freshwater inputs and terrestrially derived fluxes of alkalinity, minerals, and nutrients5,39,40. Additionally, given the diversity of mollusk responses to ocean acidification41, a mechanistic understanding of shell formation in this bivalve might help explain divergent responses across mollusk taxa in a manner similar to frameworks constructed for calcifying phytoplankton20,42. Our examination employed wide ranges of each of the four traditionally measured carbonate system parameters of seawater (dissolved inorganic carbon, total alkalinity, pCO2, and pH) (Supplementary Fig. S1). We conducted 324 separate incubations to assess the calcification responses to the three potential carbon sources available to organisms ([HCO3−], [CO32−], [CO2]). The incubations also allowed for investigating effects of pH, which influences the action of ion pumps and channels governing movement of Ca2+ and/or bicarbonate to or from precipitation sites14. We additionally varied dissolved calcium concentrations to assist in isolating co-varying factors, and we tested responses to salinity. We quantified calcification rates for all incubations using alkalinity anomaly techniques corrected for changes in excreted ammonia43. We conducted separate incubations of de-fleshed mussel shells to quantify abiotic dissolution, thereby distinguishing net from gross calcification (net calcification = gross calcification minus dissolution; Supplementary Fig. S2). Calcification rates were normalized by g0.72 to account for the allometric relationship between biomass (g) and calcification determined for this study (Supplementary Fig. S3).

Results

Neither Ω nor SIR fully explain calcification

As noted, considerable attention has focused on Ω as the dominant driver of calcification in a variety of species. A smaller set of studies emphasizes the capacity of SIR to operate analogously as a good mathematical predictor17,18,21,44. In contrast to both of these models, however, we found that mussel gross calcification responded only modestly to either Ω (Fig. 3a) or SIR (Fig. 3b). Instead, higher calcification rates associated strongly with elevations in [HCO3−]. This pattern emerged clearly in incubations where we varied [Ca2+] for a given [HCO3−], or where we varied bicarbonate concentrations for a given Ω or SIR. In all trials, we corrected for effects of abiotic dissolution (Supplementary Fig. S2); however, we also observed evidence of residual shell loss even after correction, suggesting that the biological activity of mussels45 or metabolism of shell-associated microbial communities46 may enhance dissolution beyond that predicted by abiotic processes alone. This dissolution occurred at the lowest Ω and was offset by elevated [Ca2+] (Fig. 3a, b), verifying that Ω can contribute to net calcification by modifying rates of shell dissolution, even when Ω is not a primary controller of gross shell formation per se12,47.

Fig. 3: Mussel gross calcification rates respond strongly to bicarbonate ion concentration.
figure 3

Mussel gross calcification rates (n = 28) respond strongly to increased bicarbonate ion concentration ([HCO3−]). However, bicarbonate does not explain all trends in gross calcification by itself. Possible additional contributors include (a) calcium carbonate saturation state (Ω), (b) joint effects of bicarbonate and proton ([H+]) concentration as indexed using the substrate inhibitor ratio (SIR), (c) carbonate ion concentration ([CO32-]), (d) proton concentration by itself, (e) carbon dioxide concentration ([CO2]), and (f) pH on the free scale, a logarithmic display of [H+]. Ambient [HCO3−] in these experiments was 1700 ± 300 µmol kg−1 (mean ± sd) while high [HCO3−] was 7110 ± 810 µmol kg−1 (mean ± sd). Calcium additions (open triangles) raised Ω by a factor of 1.27 to 4.60. The calcium additions mitigated the strongest negative calcification rates (no rates below −3 µmol CaCO3 h−1 g-0.72 on the y-axis for the open triangles, where g-0.72 accounts for relationship between mussel size and calcification rate; see Supplementary Materials), consistent with slowed dissolution of existing shell material.

Full size image

Considering a dual-parameter model of calcification

HCO3 − is a core driver

The inability of either saturation state (Ω) or the substrate inhibitor ratio (SIR) to fully describe gross calcification in mussels indicates that neither are adequate as single-parameter predictors. Under typical environmental conditions, bicarbonate ion concentration exhibits the least collinearity with other parameters of the seawater carbonate system48 (Fig. 1) and emerges unambiguously as one controlling factor for calcification (Fig. 3; Supplementary Fig. S1). A primary role for bicarbonate is not surprising; this is well-supported by theory and by the known existence of [HCO3−] transporters in a variety of taxa14,49 and is consistent with studies of other mussels under conditions where inorganic carbon is limiting18. At the same time, bicarbonate by itself incompletely explains gross calcification rates, in that coincident effects of at least one additional factor are apparent. This point is visible in various panels of Fig. 3 that show gross calcification rate depending not just on [HCO3−], but also changing with x-axis position (see also Supplementary Fig. S1).

Unfortunately, it is not straightforward to isolate which candidate parameters along the x-axes of Fig. 3 might join with [HCO3−] to dictate gross calcification in M. californianus. As a starting point, we focus on gross calcification models that depend on two driving parameters, as opposed to more complex models involving three or more governing factors. This approach adheres to the principle of parsimony—the concept that simpler models are more probable—and aligns better with the limited degrees of freedom in the carbonate system equilibria. The list of biologically justifiable dual-parameter options includes: (1) [HCO3−] and pH (or [H+]), (2) [HCO3−] and Ω, (3) [HCO3−] and CO32−, (4) [HCO3−] and [CO2]. We emphasize that although these driver pairs all relate to gross calcification, Ω can also modulate net calcification through effects on dissolution.

The similarity across panels a–d in Fig. 3 is striking but also expected. As is well known, various parameters in the seawater carbonate system co-vary, albeit nonlinearly. Prior studies have emphasized this fact when demonstrating, for example, why Ω and SIR vary closely. In the case of these two parameters17, their tight relationship derives from definitions of the second dissociation constant of the carbonate system (K2*) and the saturation state of calcium carbonate, respectively:

$${{K}_{2}}^{ * }=[{{{{{{\rm{CO}}}}}}_{3}}^{2-}] [{{{{{\rm{H}}}}}}^{+}]/[{{{{{{{\rm{HCO}}}}}}}_{3}}^{-}]$$
(1)
$$\Omega =[{{{{{{\rm{Ca}}}}}}}^{2+}][{{{{{{{\rm{CO}}}}}}}_{3}}^{2-}]/{{{{{{{\rm{K}}}}}}}_{{{{{{\rm{sp}}}}}}}}^{* }$$
(2)

where Ksp* is the stoichiometric solubility product constant. Combining these expressions yields

$$\Omega =[{{{{{{\rm{Ca}}}}}}}^{2+}]({{{{{{{\rm{K}}}}}}}_{2}}^{* }/{{{{{{{\rm{K}}}}}}}_{{{{{{\rm{sp}}}}}}}}^{* })[{{{{{{{\rm{HCO}}}}}}}_{3}}^{-}]/[{{{{{{\rm{H}}}}}}}^{+}]$$
(3)

Given that calcium concentrations vary little across many ocean environments, and that K2* and Ksp* depend only on temperature, salinity, and pressure, it is apparent that SIR (=[HCO3−]/[H+]) is proportional to Ω, in the absence of temperature- or salinity-driven seasonal variations at regional scales5 such as those depicted in Fig. 2, or large pressure differences across depth. Likewise, inspection of Eq. 1 indicates that [CO32−] is proportional to SIR (with K2* as a constant), and that [CO32−] and [H+] are inversely related if bicarbonate concentrations are held constant, as is the case across each of the green and orange data sets in Fig. 3. Proton concentrations are in turn, connected to [CO2] through the first dissociation constant

$${{{{{{{\rm{K}}}}}}}_{1}}^{* }=[{{{{{{{\rm{HCO}}}}}}}_{3}}^{-}][{{{{{{\rm{H}}}}}}}^{+}]/[{{{{{{\rm{CO}}}}}}}_{2}]$$
(4)

such that for fixed bicarbonate levels, [H+] and [CO2] are proportional, as K1* varies only with temperature, salinity, and pressure. Therefore, although Fig. 3 superficially implies that gross calcification might depend on bicarbonate plus as many as five additional parameters (noting that H+ and pH refer to the same chemical property), the same patterns would originate from effects of bicarbonate plus one other factor. These interrelationships are of course recognized across the fields of chemical oceanography and ocean acidification research, but are sometimes not displayed nor discussed as explicitly as would be ideal for inferring cause and effect.

Evidence against a HCO3 − and Ω model

Some of the difficulties with distinguishing among possible calcification drivers of Fig. 3 can be addressed experimentally. In our study, we disentangled effects of saturation state from the rest of the carbonate system by manipulating seawater calcium concentrations. Examining the relationship between Ω and pH, for example, it is clear from Eq. 3 that for fixed calcium and bicarbonate concentrations and for constant temperature, salinity, and pressure, Ω and [H+] vary inversely as

$$\Omega ={{{{{{\rm{C}}}}}}}_{0}/[{{{{{{\rm{H}}}}}}}^{+}]$$
(5)

where C0 = [Ca2+](K2*/Ksp*)[HCO3−]. Alternatively, noting that pH = −log [H+], and taking the base-10 logarithm of both sides of Eq. 5, reveals a linear relationship between pH and log Ω

$$\log \Omega ={{{{{{\rm{C}}}}}}}_{1}+{{{{{\rm{pH}}}}}}$$
(6)

where C1 = log(C0), which increases with calcium or bicarbonate concentration. Equation 6 highlights that gross calcification rates that rise with Ω could be due to higher pH, or vice versa; effects of one readily masquerade as the other. Importantly, however, [Ca2+] negligibly influences carbonate equilibria but increases saturation state and, thus, any rate-mediated control over calcification. As a result, if gross calcification remains unchanged with higher calcium, Ω cannot be a driver.

We did not observe an effect of calcium carbonate saturation state on gross calcification rates in our data (Fig. 4). Elevated calcium and, consequently, higher Ω at a given [HCO3−] (4100 ± 100 µmol kg−1, mean ± sd, n = 50) leaves calcification unchanged (Fig. 4a) whereas gross calcification was strongly correlated with the other carbonate system parameter (Fig. 4b–f).

Fig. 4: Mussel gross calcification rate not explained well by saturation state.
figure 4

At a fixed bicarbonate ion concentration ([HCO3−]), gross calcification rates are not explained by (a) calcium carbonate saturation state (Ωaragonite, F1,47 = 0.019, p = 0.891). Mussels exposed to seawater with elevated calcium ion concentrations ([Ca2+] additions that doubled Ωaragonite) do not show higher calcification rates relative to those in ambient seawater (compare triangles versus circles of a given color, respectively). Possible additional explainers including (b) SIR, (c) [CO32-], (d) [H+], (e) [CO2], and (f) pH are shown and are discussed in the text. For these incubations, [HCO3−] = 4100 ± 100 µmol kg−1 (mean ± sd, n = 50). Error bars in the y direction represent the standard error of calcification rates in each group while error bars in the x direction represent the standard deviation among the incubations in each treatment.

Full size image

Reasons against a HCO3 − and CO3 2− model

We can also discount a gross calcification construct based on bicarbonate and carbonate for adult M. californianus. In situations where an organism’s precipitation site sits in direct contact with unmodified seawater, which is not the case for adult mussels, it is reasonable to anticipate that CO32− could contribute directly to gross calcification, possibly through carbonate’s effect on saturation state. In larval bivalves for instance, the shell is formed external to most body tissues and mineralization may proceed in a location bathed by surrounding seawater9. In other taxa, such as some calcifying phytoplankton, bulk seawater can be transported to the calcifying site50, again provisioning CO32− for calcification. However, in cases where calcification takes place within an isolated fluid volume, a strong role for carbonate ions becomes more difficult to envision. Even if unmodified seawater were present at the calcifying site, HCO3− would also be present and at concentrations much greater than CO32− (e.g., at a 7.5 pH, [HCO3−]/[CO32−] > 50 and this ratio remains greater than 5 at 8.5 pH). Additionally, calcium carbonate precipitation in adult bivalves likely originates within hemocytes that transport the initial crystals through the extrapallial fluid to the mantle tissue where the shell is being formed, a process requiring ion transport across membranes51. Further, oxygen isotope data from M. californianus shells suggest strong biological control of the extrapallial fluid pH52 further emphasizing that these mussels probably do not rely on bulk seawater transported directly to the calcifying space. These details, in addition to the lack of identified CO32− ion transporters14,21, lead us to infer that carbonate does not contribute a major source of inorganic carbon in adult California mussels.

Reasons against HCO3 − and CO2

Dual control of gross calcification by means of a HCO3− and CO2 model similarly appears improbable. Prior research demonstrates little role for dissolved CO2 (also discussed as pCO2 in the literature) in shell precipitation by marine mollusks. Isotope analyses indicate that inorganic carbon in these organisms tends to originate from ambient seawater rather than from CO2 produced through respiration13. Reverse radioisotope labeling experiments with other bivalve species also discount reliance on respiratory carbon, showing its use varies little (<3%) as external seawater [CO2] increases53. Were respiratory carbon a major source, it should comprise a greater fraction of shell carbon in cases where its diffusion from the precipitation site is impeded by higher external [CO2]. Moreover, the pH of the calcifying fluid is likely lower than that of the surrounding seawater (a pattern observed in a congener, Mytilus edulis27), which would oppose passive diffusion of CO2 to the calcification site. A final argument against [CO2] working in conjunction with bicarbonate to drive gross calcification in California mussels is that patterns of Figs. 3e and 4e are opposite to what one would expect if CO2 were operating as a substrate for CaCO3 formation. If the latter scenario were occurring, then higher CO2 concentrations should associate with greater rates of calcification rather than the reverse. The pattern we observe whereby increased CO2 reduces calcification (Figs. 3e,4e) is consistent with impacts on acid-base regulation and a need to control pH in the calcifying space54 as opposed to inhibition by CO2 per se.

Calcification is explained best by a HCO3 − and pH model

Each of the above lines of evidence and reasoning lead us to conclude that HCO3− and pH (or [H+]) together control calcification in adult California mussels. To test this, in addition to our core incubation experiments (Figs. 3 and 4), we also verified that observed trends held across a more comprehensive array of potential calcification drivers and for an extensive number of mussels (n = 324 total). Over this full dataset, bicarbonate and pH consistently exhibited simultaneous and independent explanatory power in describing gross calcification rates (Fig. 5). The elaborated decoupling efforts also demonstrated that neither salinity nor [Ca2+] were good predictors of calcification by themselves, aside from the latter’s ability to mitigate against dissolution when saturation state declined below one (Fig. 3; see also Supplementary Figs. S4–S5). Results indicated a minor trend of larger mussels exhibiting lower biomass-normalized calcification rates (Table 1), possibly indicating that smaller mussels may prioritize calcification over other physiological processes to a greater extent than larger mussels, as has been observed in other bivalves55. Calcification rates additionally exhibited a saturating trend as [HCO3−] rose (linear trend after log-transforming [HCO3−]; Fig. 5, Table 1), a pattern consistent with the operation of concentrating mechanisms for bicarbonate at the site of calcification, potentially via transport across cell membranes32,33. Saturation of a HCO3– uptake pathway has been modeled previously in coccolithophores with Michaelis-Menten kinetics, while also allowing calcification rates to be independently inhibited by elevated [H+]20. A Michaelis-Menten representation appears less successful in reproducing calcification rates of the mussels in our experimental system (Supplementary Fig. S6), although further work on this topic is warranted.

Fig. 5: Mussel gross calcification rates depend on multiple parameters.
figure 5

Bicarbonate ion concentration ([HCO3−]) is a driver of mussel gross calcification rates. Here, [HCO3−] is displayed on a log scale to account for the saturation of calcification rates with rising [HCO3−]. Colors are used to highlight potential drivers that may also contribute to gross calcification rates. Candidate drivers include (a) carbonate ion (CO32-) concentration in µmol kg−1, (b) the substrate to inhibitor ratio (SIR), (c) carbon dioxide (CO2) concentration in µmol kg−1, and (d) pH on the free scale. Although it appears that any of these factors could drive gross calcification, this situation is unlikely due to the lack of CO32- transporters and the lack of parsimony if SIR were a driver. In the case of [CO2], were this compound operating as a substrate for calcification, higher levels of carbon dioxide (blue color) should lead to higher rates of gross calcification. Assessing these trends together, gross calcification appears to be driven by the combined effects of [HCO3−] and pH. See text for more details. All colors assigned on a logarithmic scale except for pH which already accounts for a non-linear effect of [H+].

Full size image
Table 1 Fixed effects from linear mixed effects models testing the contributions of each parameter to gross calcification rates (µmol CaCO3 hr−1 g-0.72) for (a) a dual parameter model considering pH and bicarbonate ion (HCO3−) concentration (marginal R2 = 0.73, conditional R2 = 0.80, AIC = 1485.19, BIC = 1507.87, residual variance = 5.208) and (b) a single-parameter model considering only the substrate to inhibitor ratio (marginal R2 = 0.60, conditional R2 = 0.73, AIC = 1568.88, BIC = 1587.79, residual variance = 6.708) across all 324 mussels
Full size table

Discussion

Results of our study indicate that, across the multiple axes of the seawater carbonate system, no single parameter fully explains calcification rates in mussels by itself. Only by accounting for the independent action of two distinct parameters, most likely [HCO3−] and pH, could a complete description of gross calcification rates be established (Fig. 5, Table 1). This dual-parameter model highlights the importance of alkalinity perturbations for M. californianus in coastal systems where freshwater inputs from rivers or rain can drive substantial alterations to bicarbonate and pH, and thus to calcification (Fig. 6). It also emphasizes the efficacy of considering relationships among multiple environmental factors (e.g., altered patterns of precipitation in combination with ocean acidification5) when attempting to accurately project consequences of rising CO2 (Fig. 6). Indeed, our findings suggest that previously observed decreases in M. californianus calcification rates associated with rainfall could arise from effects of altered [HCO3−] as opposed to physiological challenges associated with lowered salinity56.

Fig. 6: Chemistry perturbations in coastal ecosystems have differing implications for mussel calcification.
figure 6

Carbonate chemistry perturbations common in coastal areas and arising due to routine ecosystem-level processes may strongly influence mussel calcification rates. Predicted calcification rates for Mytilus californianus across a range of pH and bicarbonate ion concentration (HCO3−). Colors indicate calcification rate in µmol h−1 g-0.72 predicted by the calcification model (Table 1). Example perturbation arrows and field data are the same as described in Fig. 1. Note that although M. californianus mussels are not found naturally in all environmental locations shown via the symbols here, a wide breadth of environmental data are included for context. Processes that alter alkalinity, such as rain or high alkalinity river input, run more perpendicular to the calcification rate contours and, thus, are projected to have stronger consequences for mussel calcification. Dashed contours represent the ratio between calcification rates predicted by independent effects of [HCO3−] and pH, and those predicted by log10(SIR) alone. For example, for an increase in CO2 to 1200 µatm, characteristic of end of century predictions for CO2 levels or an environment dominated by respiration (Resp perturbation arrow), the model that considers independent actions of [HCO3−] and pHfree would predict calcification rates 1.2 times faster than rates predicted from a model that uses their mathematical ratio. Differences between the two models are most pronounced when considering CO2 perturbations like respiration, OA, or photosynthesis.

Full size image

Much as results of modifying [Ca2+] (Fig. 4) demonstrate that Ω cannot unilaterally explain calcification in California mussels, independent effects of HCO3− and pH in our study showcase a limitation of using SIR as a single-parameter predictor. Both the independent HCO3− and pH framework, as well as the mathematical ratio [HCO3−]/[H+], emphasize the importance of HCO3− as a calcification substrate. Both also acknowledge the need to overcome biochemical inhibition from waste protons (i.e., H+ must be discarded when bicarbonate is utilized in forming CaCO3). However, the ratio of [HCO3−]/[H+] is only capable of accurately predicting calcification across a range of carbonate system conditions if effects of high bicarbonate concentrations duplicate effects of low proton concentrations, and vice versa. This situation is possible only if a given percent increase in bicarbonate has identical consequences for calcification as an equivalent percent decrease in [H+]. Implicit to such a constraint is that inhibition would have to occur in an exact inverse fashion to substrate acquisition, a narrow pattern most easily rationalized if both processes proceeded along the same biochemical pathway. Such a scenario is unlikely given that inorganic carbonate acquisition and pH regulation probably utilize distinct transport mechanisms34.

As was briefly alluded to above, our findings also have implications for contextualizing prior experiments concerning Ω-attributed calcification in bivalves, including among larval stages of M. californianus. In these earlier studies, some of which also employed a decoupled carbonate system, shell formation appeared to be governed primarily, if not exclusively, by Ω (refs. 8,11,57.). This trend appears inconsistent with results shown here (Figs. 2 and 3). However, larval stages rely on more rapid shell formation than adults, with the mineralization site likely in direct contact with external seawater9. Such high calcification rates lacking a biologically controlled calcifying fluid are believed to be limited by mineral precipitation kinetics7,8,9,10,11 that may be less applicable to post-larval life stages that do not calcify as rapidly. By the benthic juvenile stage, it is much more likely that limitations to inorganic carbon uptake of bicarbonate occur, as HCO3− exerts much stronger control over calcification10,18. Whether larval versus juvenile or adult differences in shell precipitation ultimately derive purely from ontogenetic shifts in calcification physiology, or whether allometric considerations across size (e.g., area-volume tradeoffs regarding the balance between precipitation and dissolution kinetics) contribute as well, deserve further study. Moreover, this difference highlights the need to avoid the uniform application of a singular model, even within a species, when attempting to predict the impacts of altered seawater chemistry on coastal ecosystems.

Comparisons across species beyond bivalves are also important but remain sparse. To our knowledge, few experiments exist in other phyla that deliberately test the contribution of multiple carbonate system parameters to calcification. Those studies that do undertake such efforts primarily test growth in photosynthetic calcifiers like corals29,30,31,35,36,37,58, coralline algae37,58, or coccolithophores20,32,33,59. However, calcification and photosynthesis can be tightly coupled through a variety of mechanisms60,61, making it difficult to assign calcification responses to a specific carbonate system parameter. Some of these other studies, nevertheless, do support the hypothesis that single parameters like Ω and SIR might oversimplify the description of calcification. For example, a study of two coral and two coralline algae species acclimated to different conditions of manipulated seawater chemistry suggests that pH and dissolved inorganic carbon (DIC) should be considered separate drivers of calcification58. Considering non-linear effects of seawater [HCO3−] that operate separately from both the inhibitory action of pH and fertilization by [CO2], also appears to help explain differences in growth across different species of coccolithophores20. Further experiments exposing non-photosynthetic calcifying species to a decoupled carbonate system are warranted to determine the ubiquity of multiple, independent drivers of calcification and to further our understanding of the breadth of calcification mechanisms present across marine groups. Interestingly, species exposed to ocean acidification conditions differ in how they interact with seawater pH27. So even if [HCO3−] and pH are drivers of calcification in other species and life stages, differences in the relative sensitivity to these two parameters will alter our predictions of how marine organisms will respond to chemical perturbations. Recognizing such variation is an important step towards generating simplified frameworks that can assist with understanding how coastal ecosystems across diverse oceanographic conditions will respond to climate change.

The likelihood that separate pathways govern inorganic carbon acquisition and proton removal also has broader relevance to ocean life. In coastal habitats and estuaries, departures in alkalinity and DIC from open-water values can alter bicarbonate and pH levels appreciably (Fig. 2). Such deviations from ambient conditions are sufficient to induce large changes in calcification rates (Fig. 6), with implications for organism performance. Even in less-dramatic scenarios where rates of CaCO3 formation might shift only modestly, accompanying perturbations to competitive or predatory relationships could induce altered species interactions that are increasingly recognized as agents of acute ecological change62. Potential cascading effects of this nature warrant attention in future work. Longer-term experiments that allow for an evaluation of acclimation responses or genetic adaptation are also needed, although the short-term (~2 h) incubations of our study are well-suited to disentangling calcification drivers over certain ecological timescales. In particular, they align well with the rapid fluctuations in carbonate chemistry typical of coastal-zone habitats, which are often driven by tidal exchanges, episodic storm events, or hourly-scale biological processes63,64.

Our data support growing evidence that single parameter models may not adequately describe calcification for many marine species. Calcification rates modeled with a dual-parameter, HCO3− and pH framework can diverge substantially from those modeled solely by a single parameter. Consider, for example, a scenario whereby dissolved seawater CO2 rises from 400 µatm to 1200 µatm. Such changes can occur as the result of local community respiration56,65 but are also consistent with end of century projections of CO2 levels66. Calcification rates predicted with a single-parameter Ω or SIR model would decline nearly 45% in such conditions, compared to a model recognizing the independent roles of [HCO3−] and pH (Fig. 6) which would only predict a 31% reduction in calcification. In both models, the decrease in gross calcification is driven by increases in [H+] that are unable to be offset by concurrent increases in [HCO3−]. Consider also a situation where active photosynthesis by autotrophs locally reduces carbon dioxide67, lowering dissolved seawater CO2 from 400 µatm to 280 µatm, also characteristic of pre-industrial CO2 levels1. An Ω or SIR model would predict approximately 11% higher calcification rates, while the dual-parameter model would predict only a 7% increase (Fig. 6). Though both models suggest that mussel calcification rates have decreased since pre-industrial times, the extent of this decrease differs.

These examples can be expanded to accommodate more complex carbonate system perturbations, as well. Both single-parameter Ω or SIR models, and the dual-parameter (independent effects of HCO3− and pH) model, predict that increased alkalinity will elevate calcification rate. However, if [CO2] increases, accompanying alkalinity modification by calcification or dissolution can either amplify or attenuate differences between the single versus dual-parameter models (see relevant process arrows in Fig. 6). In extreme cases, the model differences can reach several-fold. This class of outcomes contrasts with those due to alkalinity modifications driven by rain or high-alkalinity river inputs where the added impact of these processes causes large changes in calcification rate per se, but less divergence between predictions across models. The major trends of these relationships moreover reverse when [CO2] is reduced rather than increased; here, perturbations due to rain and rivers drive stronger differences between the single- and dual-parameter models than carbonate system modifications driven by calcification or dissolution. Therefore, understanding the mechanisms by which calcifying organisms interact with chemical surroundings is crucial to projecting responses to future environmental changes.

Conclusions

Single-parameter models of calcification can oversimplify the carbonate system, challenging our ability to predict the impacts of global change on coastal marine communities. We argue here for a broader adoption of a multi-parameter approach when attempting to understand calcification rates, even in species lacking photosynthesis. An independent bicarbonate and pH model, which our data support for adult Mytilus californianus, expands our understanding of how calcifiers respond to variable seawater chemistries characteristic of coastal environments, and encourages further exploration of the mechanistic basis for those responses in the context of global ocean change. These findings thus have substantive implications for appropriate management and protection of coastal ecosystems against impacts of ocean change.

Methods

Mussel collections

We gathered naturally settled, adult California mussels (M. californianus between 30 and 80 mm in maximum shell length) by hand from the mid-intertidal zone of Carmet Beach, along the northern California coast (38°22'17.0“N 123°04'33.8”W). We confined our sampling to a single population to minimize variation in physiologically plastic traits that can vary spatially68. We cleaned mussels of all epibionts and external byssal threads, then transported them in buckets (<0.5 h transit time) to Bodega Marine Laboratory, where we acclimated individuals for seven days in flow-through seawater tables prior to subsequent experiments. We conducted collections and subsequent experiments nine times over a two-year period from January 2020 to April 2022.

Incubations

We determined effects of carbonate chemistry on shell formation in Mytilus californianus by observing calcification rates in 324 adult mussel individuals, each exposed to a unique carbonate chemistry treatment. We conducted incubations in the dark, in a temperature-controlled room, where room temperature was set to match ambient seawater temperatures during the collection and subsequent acclimation (13.5 °C for all experimental modules except in April 2022 when it was set to 11.5 °C). Experiments consisted of placing a single mussel into approximately 0.85 kg seawater in a 1 L airtight incubation vessel for between 0.8 and 2.7 h. Variability in incubation duration nominally reflected variability in mussel size to allow for a sufficient alkalinity change without driving oxygen below 85% of the starting value. Prior to beginning an incubation—starting with seawater near zero alkalinity and zero dissolved inorganic carbon (DIC)—we added a DIC stock solution, along with a solution of either hydrochloric acid or sodium hydroxide, distilled water, or a calcium stock solution to achieve specific DIC, alkalinity, salinity, or calcium treatments, respectively (see Chemistry Manipulations section below). Note that all chemical manipulations produced seawater conditions that, although different from ambient values, are comparable to conditions observed in the coastal ocean or in tidepools within our region63,64. Concomitant with alkalinity assays, we also measured respiration rates of each individual mussel, quantified from the change in seawater [O2] during the incubation, as measured by a Presens Microx 4 oxygen sensor.

We calculated net calcification rates with the ammonia-corrected alkalinity anomaly technique43, divided by incubation duration and mussel dry tissue mass raised by a factor of 0.72 (described below in Allometric scaling of calcification). The alkalinity anomaly technique builds on the observation that precipitation of CaCO3 results in an equivalent reduction in seawater [CO32−] (or reduction of [HCO3−] followed by an increase in [H+]) which contributes two equivalents of total alkalinity—simultaneous production of ammonia is the major metabolic process in mussels that can obscure this and its signal must be removed43. Following the incubation, we dissected each mussel and dried it at 60 °C for at least 24 h to obtain the dry tissue mass (excluding byssal threads) and dry shell mass of each individual mussel. We conducted additional incubations (n = 87, between 3 and 9 per experiment day) without mussels throughout the trials as experimental blanks to determine background changes in alkalinity (Supplementary Fig. S7). We excluded from our analysis any experimental days where background alkalinity changes exceeded 5 µmol kg−1. The mean of the absolute values of alkalinity change during the incubations of these experimental blanks was 1.3 ± 1.2 µmol kg−1 (n = 72).

Abiotic dissolution signal

We used separate incubations with de-fleshed mussel shells to quantify rates of abiotic dissolution, and we employed these dissolution rates to correct the alkalinity anomaly data to estimate gross calcification rates (gross calcification = net calcification + dissolution). We dried and bleached shells (n = 60) originating from live mussels at Carmet Beach, CA, and used 7.5% sodium hypochlorite to remove excess tissue and microbial communities, before incubating them in an analogous fashion to the calcification trials (Supplementary Fig. S4). We fit dissolution rate data, plotted against calcium carbonate saturation state, with an Arrhenius-derived dissolution equation of the form y = b0 – b1*ea*Ω where y is the measured dissolution rate, b0 is the asymptotic dissolution rate, b1 is the y-intercept, and a is the rate of approaching the asymptote12. We applied the dissolution corrections prior to normalization by dry tissue weight (see Allometric scaling of calcification below).

Allometric scaling of calcification

To account for calcification scaling allometrically with animal size, we determined the relationship between mussel tissue mass and calcification rate by incubating 26 different mussels spanning 24 mm to 67 mm in shell length and 0.07 g to 2.55 g of dry tissue in ambient, unmanipulated seawater (Supplementary Fig. S3, F1,24 = 61.76, R2 = 0.71, p < 0.001). This procedure yielded a scaling exponent of 0.72 and we therefore divided calcification rates of the experimental mussels by g0.72.

Chemistry manipulations

During incubations, we altered the seawater of each incubation vessel to achieve a specific combination of alkalinity, dissolved inorganic carbon (DIC), and [Ca2+] to decouple the parameters of the carbonate system. We adapted and expanded this approach from previously published methods11. To attain each chemical treatment, we first acidified 20 L sumps filled with filtered seawater with hydrochloric acid (HCl) to reduce the alkalinity to near zero and drive the DIC equilibrium towards CO2. We then bubbled this acidified water vigorously with air to off-gas the CO2. For each incubation vessel, we added seawater from these sumps to a 1 L mixing vessel. Then, we re-added DIC in the form of sodium bicarbonate and sodium carbonate to reach target levels and adjusted the final alkalinity with either HCl or sodium hydroxide (NaOH). We mixed the altered seawater vigorously before subsampling it, collecting approximately 150 mL for alkalinity and 6 mL for spectrophotometric pH analyses (see Chemical Analyses section below). We added to the incubation vessel the remaining 850 mL, along with an experimental mussel individual, after recording the water mass. We chose treatments to span the range of carbonate system conditions present in natural coastal environments along the west coast of the U.S., especially in association with locations with riverine input and/or within intertidal rock pools where biological activity strongly alters the carbonate system63,64.

Experimental treatment targets

We compared effects on calcification of the composite parameters, Ω and SIR, to the effect of [HCO3−] in three experimental trials (n = 28 incubations total). In the first trial, used to determine baseline responses to carbonate levels, we exposed mussels (n = 7) to seawater spanning a range of [CO3−] (4.31 <[CO3−] < 246 µmol kg−1) with [HCO3−] fixed at 1700 ± 300 µmol kg−1 (mean ± sd). In the second trial, we incubated mussels (n = 8) under elevated [Ca2+] (12 < [Ca2+] < 50 mmol kg−1), achieved via the addition of a concentrated solution of CaCl2, and the same bicarbonate level to independently alter Ω without changing carbonate ion concentration; these manipulations raised Ω by a factor of 1.27 to 4.60 compared to seawater with equivalent carbonate chemistry and ambient calcium. In the third trial, we examined calcification effects of bicarbonate by incubating mussels (n = 13) over a similar range of Ω and natural [Ca2+], but with elevated [HCO3−] = 7110 ± 810 µmol kg−1 (mean ± sd). Results of these trials appear in Fig. 3.

To isolate whether pH or Ω accompanies bicarbonate as a second driver of calcification, we incubated mussels (n = 50) under combinations of high or low pH, and elevated or ambient [Ca2+]. We filled two large sumps with filtered seawater and sparged them with ambient air. To one sump, we added enough solid CaCl2 to raise [Ca2+] by a factor of ~2.5. To the other, we added solid NaCl at 1.5 times the molar equivalent of the added CaCl2 so that the salinity increases were equivalent. We then added DIC stock and NaOH as described above to generate treatments of low pH and ambient [Ca2+] (7.74 ± 0.05 pH and 10.7 ± 0.1 mmol Ca2+, mean ± sd, n = 14), low pH and elevated [Ca2+] (7.69 ± 0.06 pH and 24.9 ± 0.4 mmol Ca2+, mean ± sd, n = 13), high pH and ambient [Ca2+] (8.03 pH ± 0.05 and 10.5 ± 0.2 mmol Ca2+, mean ± sd, n = 13), and high pH and elevated [Ca2+] (8.00 ± 0.02 pH and 23.7 ± 0.4 mmol Ca2+, mean ± sd, n = 10). We held bicarbonate ion concentration constant during these incubations at 4100 ± 100 µmol kg−1 (mean ± sd, n = 50). Results of these trials appear in Fig. 4.

Remaining incubations (n = 246) focused on modifying carbonate chemistry, at ambient salinity and [Ca2+], to expand the breadth of values across other relevant parameters, spanning variation in pH (6.63 < pH < 8.84), [HCO3−] (280 < [HCO3−] < 8830 µmol kg−1), and [CO2] (1 < [CO2] < 536 µmol kg−1, corresponding to pCO2 between 26 and 13,465 µatm). In addition, because changes in carbonate chemistry can associate with processes that change seawater salinity22,69, we adjusted salinity in another set (n = 41) of incubations to 30.46 ± 0.64 (mean ± sd, n = 21) and 26.85 ± 0.37 (mean ± sd, n = 20). We altered salinity by diluting seawater with distilled water before adding the stock DIC and alkalinity solutions. Experiments to alter salinity also altered [Ca2+] so we conducted additional experiments (n = 55) where Ca2+ was replaced by adding an aliquot of either 1 molar or 5 molar calcium chloride to independently test effects of salinity and [Ca2+]. In these latter experiments, salinity ranged between 18.68 and 33.23 and [Ca2+] ranged between 5 and 15 mmol kg−1 (Supplementary Fig. S4). Natural salinities in our region are typically between 33.65 and 34.36, and rarely fall below 2570. Mussels exposed to the lowest salinity treatments did not calcify or respire, and data from these runs were therefore excluded from further analysis (n = 11). Although Ca2+ and Na+ additions (from CaCl2 and NaCl, respectively) can alter the salinity measured by conductivity, these perturbations were minimal (<25 mmol kg−1) and so their influence on ionic strength and equilibrium constants48 was ignored.

Chemical analyses

We quantified chemical changes during the incubations by differencing measurements taken immediately before and after each trial. We defined treatment conditions as the average value between the starting and ending chemistries (variation typically <2% of starting value), and calculated rates as the differences in those values divided by incubation duration. We measured oxygen and temperature with a PreSens Microx 4 micro-optode whose factory calibration was verified in air-equilibrated seawater. We replaced or recalibrated according to manufacturer’s instructions sensors deviating more than 5% saturation from the 100% value. We measured seawater pH with either a custom pH meter using a Honeywell DuraFET probe (pre July 2020) or a Horiba Laqua PC1100 instrument. We calibrated both sensors daily on the total scale with simultaneous spectrophotometric measurements of pH at incubation temperature; we conducted these calibrations in a temperature controlled room using the pH sensitive dye, m-cresol purple71, immediately after sample collection. We analyzed water samples collected for alkalinity on a Metrohm 855 Titrosampler within 24 h of collection where acid concentration was determined daily using certified reference materials from the laboratory of Dr. Andrew Dickson (Scripps Institute of Oceanography). Due to volume limitations, we analyzed starting alkalinity in duplicate (average standard deviation of samples = 4.46 µmol kg−1), while we analyzed alkalinity at the end of the incubations in triplicate (average standard deviation of samples = 1.87 µmol kg−1). We determined ammonia with a salicylate spectrophotometric assay72 to correct for changes in alkalinity due to mussel waste excretion43. We determined initial ammonia daily from each sump, while we determined ending ammonia with water collected from each incubation vessel. No other nutrients (nitrate, phosphate, etc) were measured and were assumed to have not changed dramatically during the incubations. We determined calcium concentrations, when measured in [Ca2+]-manipulation experiments, with an Oakton combination calcium ion-selective electrode calibrated with samples analyzed for [Ca2+] by titration with EGTA73. We immediately froze water samples collected for ammonia and calcium determinations until later analysis. We measured the salinity of each sump with a Yellow Springs Instruments (YSI) ProPlus conductivity probe. After July 2020, we measured salinity of individual treatments with a Horiba Laqua PC1100 conductivity probe. We calibrated both conductivity probes with YSI conductivity standards (50 mS cm−1).

Statistical analyses

We performed all computations with R statistical software, version 4.1.0. We performed carbonate system calculations using the package seacarb74, using equilibrium constants from ref. 75. We computed linear mixed models using the lmer function in the lmertest76 package in R and focused on assessing likely candidate parameters (see text) as fixed factors, and mussel collection date as a random intercept to account for natural seasonal differences between cohorts. Conditional R2 was calculated with the package MuMIn77. We determined parameters for non-linear fits employed to model dissolution rates by minimizing the sum of squares of model residuals using the optim function. Colors for plots were chosen from color palettes in the cmocean78 package in R.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.