Universal response pattern of phytoplankton growth rates to increasing CO 2

Phytoplankton growth rate is a key variable controlling species succession and ecosystem structure throughout the surface ocean. Carbonate chemistry conditions are known to inﬂuence phytoplankton growth rates but there is no conceptual framework allowing us to compare growth rate responses across taxa. Here we analyse the literature to show that phytoplankton growth rates follow an optimum curve response pattern whenever the tested species is exposed to a sufﬁciently large gradient in proton (H + ) concentrations. Based on previous ﬁndings with coccolithophores and diatoms, we argue that this ‘universal reaction norm’ is shaped bythestimulatinginﬂuenceofincreasinginorganiccarbonsubstrate (leftside ofthe optimum) andtheinhibiting inﬂuence of increase H + (right side of the optimum). We envisage that exploration of carbonate chemistry-dependent optimum curves as a default experimental approach will boost our mechanistic understanding of phytoplankton responses to ocean acidiﬁcation, like temperature curves have already boosted our mechanistic understanding to global warming. not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Ofﬁce.


Introduction
Phytoplankton are responsible for about 50% of primary production on Earth (Field et al., 1998). They utilize CO 2 and bicarbonate (HCO 3 − ) as the dissolved inorganic carbon (DIC) source for photosynthetic carbon fixation (Burkhardt et al., 2001;Rost et al., 2003). Carbon fixation rates are limited by the slow catalytic rate of the RubisCO, the enzyme which binds CO 2 (Raven & Johnston, 1991). Phytoplankton hence invest energy in cellular carbon concentrating mechanisms (CCMs), which increase the substrate (i.e. CO 2 ) concentration in the vicinity of the RubisCO enzyme and drive the carboxylation reaction. However, the cytosolic membrane is highly permeable to CO 2 . Consequently, the higher the CO 2 concentration in the seawater outside the cell, the less CO 2 is lost by diffusion from the cell, thereby increasing net CCM efficiency (Hopkinson et al., 2011).
In early studies, pH was manipulated to study the sensitivity of plankton growth to increasing inorganic carbon limitation at high pH/low CO 2 (Swift & Taylor, 1966;e.g. Goldman et al., 1982;Hansen, 2002;Søderberg & Hansen, 2007). This revealed underlying differences in carbon concentrating efficiency (Reinfelder, 2011) and RubisCO specificity between phytoplankton species (Tortell, 2000). More recently, studies have focused at the other end of the pH spectrum with low pH/high CO 2 scenarios to test the sensitivity of phytoplankton to ocean acidification (OA). At high CO 2 phytoplankton growth rates can decline in some species (e.g. Berge et al., 2010;McMinn et al., 2014;Sett et al., 2014), presumably due to proton (H + ) inhibition as a consequence of the increased energetic costs of maintaining pH homeostasis in the cell (Bach et al., 2011). Proton inhibition has been directly confirmed for the growth response in the coccolithophore, Emiliania huxleyi (Bach et al., 2011) and three diatom species (Shi et al., 2019), and implied for dinoflagellates (Berge et al., 2010). Michaelis-Menten (or Monod) kinetics are commonly used to parameterize the phytoplankton growth response to CO 2 , from the species-specific physiological response (Riebesell et al., 1993;Hutchins et al., 2013) up to global scales in biogeochemical models, for example, Dutkiewicz et al. (2015). However, growth rate inhibition by H + is not considered in the commonly applied Monod model kinetics. Often a plateau in growth is assumed once a critical pH is reached as has, for example, been observed within a confined range of H + for selected diazotrophic cyanobacteria taxa . Investigation of the optimum curve response is also potentially constrained by the employed study design. In many studies, CO 2 /pH treatments are usually selected within an OA relevant range (180-780 µatm), as suggested for OA scenario testing (Barry et al., 2011). This provides specific information on selected OA scenarios tested, but limits how mechanistic information on phytoplankton growth responses could be parameterized into biogeochemical models. An understanding of broader response patterns outside of the specific OA scenarios studied can, however, be very helpful for accurate model implementation (e.g. Muller & Nisbet, 2014).

Description of the phytoplankton response to CO 2
Here, we tested the applicability of optimum growth curve response pattern found in calcifying phytoplankton (Swift & Taylor, 1966;Langer et al., 2006;Bach et al., 2011) to a broader range of phytoplankton taxa, primarily using the growing body of literature on the response of phytoplankton to changes in seawater carbonate chemistry. This may also reveal if CO 2 optima in growth rates of individual phytoplankton species are systematically different between taxa, for example, coccolithophores and diatoms as this may have implications for shifts in phytoplankton community structure under changing seawater pH.
Data was compiled from literature searches for available raw datasets using the PANGAEA data depository collection from the Ocean Acidification International Coordination Centre (OA-ICC, Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation www.newphytologist.com Forum Yang et al., 2016) and citing literature available up until June 2019. An additional unpublished dataset using the diazotrophic cyanobacteria, Crocosphaera watsonii, from our laboratory was also included as it fitted the required criteria detailed in the following. Datasets on phytoplankton were selected for model fitting based on four criteria. Firstly, five or more distinct pH/CO 2 treatments are provided. Secondly, the range of pH/pCO 2 treatments is wider than OA relevant range (pH F (free scale) < 7.4 or > 8.2; pCO 2 < 200 µatm or > 1200 µatm; OA relevant range pCO 2 = 400-1000 µatm). Thirdly, growth rates are reported for individual strains. Finally, pH was reported or alternatively data was available to calculate [H + ]. An overview on all studies considered, culturing information about Crocosphaera watsonii, and those studies selected for model fitting for each phytoplankton group is provided in Supporting Information Methods S1 & S2 and Tables S1-S3. Due to the over-proportionally large representation of coccolithophore genotypes fitting these criteria (mainly E. huxleyi strains), only five coccolithophore species were considered in our figures.
All growth rates (µ) were normalized to the maximum measured rate for each taxa in each study and then fitted to a modified Monod model (Eqn 1) by Han & Levenspiel (1988). This modified equation considers the substrate concentration as a rate limiting factor below a certain concentration (i.e. Monod kinetics, Monod (1949)) and as an inhibitor above this critical concentration.
Here, k is the reaction rate constant or maximum possible growth rate (i.e. V max equivalent), [cell] is the cell concentration, [H + ]* is the critical inhibitor concentration above which no growth occurs, n and m are fitting constants which if ≠ 0 indicate inhibition behaviour of H + , and C M is the Monod constant, that is, the substrate concentration below which no growth occurs.
We assume that at low CO 2 , the optimum curve response is driven by the stimulating influence of the substrate (CO 2 ). Evidence to show CO 2 and not pH is dominant driver of growth rates at low CO 2 is compiled in the (Figs S1-S4; Notes S1). At high CO 2 , the inhibitory effect of H + controls the growth response (Bach et al., 2015;Kottmeier et al., 2016). Under constant total alkalinity (OA scenario where TA = 2300 µmol l −1 , T = 15°C), [H + ] and CO 2 concentration are linearly correlated, provided that pH F is below 8.3, with small deviations from linearity above pH F 8.3 (Fig. 1). Due to the quasi-linear relationship between CO 2 and H + , we could still extract information on the general response pattern, without any necessary approximation of CO 2 from the datasets where only pH was reported. Hence, for consistency across all datasets used and mathematical simplicity, the optimum curves were fitted only with [H + ] as this still implicitly considers inorganic carbon substrate supporting phytoplankton growth due to the correlation with CO 2 (Fig. 1). Nevertheless, for orientation we add the approximate OA relevant pCO 2 range to Fig. 1 as grey shaded area.

Evidence for an optimum curve response pattern across diverse phytoplankton taxa
In general, an optimum curve response pattern was observed in growth rates across the species in this study. Growth rates increase with increasing CO 2 (or H + using the quasi linear CO 2 to H + correlation detailed in the previous section) but taper off at high CO 2 concentrations (Fig. 2). One exception was Trichodesmium spp., where the metabolic limits for this taxon did not appear to be reached within the experimental H + range studied (equivalent to c. 1500 µatm pCO 2 , Hutchins et al., 2015). Although a large number of studies were found in our literature search, in particular for coccolithophores and diatoms (52 and 122 responses of individual species or strains found, respectively), only 28 datasets out of 251 across all species and groups fulfilled all selection criteria (see Table S3a-h). Comparatively high interest in the response of calcifying haptophytes and diatoms compared to other phytoplankton groups is likely due to their biomineralization processes and prominent bloom formation in the ocean.
A model fit was made with the data in Fig. 2, which provides an idea as to how the taxa sensitivity can be illustrated by the Monod constant (C M ) for the lower end of substrate availability, and by the critical substrate concentration [H + ]* for the higher end of substrate concentrations where inhibition occurs. Alternatively, the sensitivity of each taxa could also be conceptualized in terms of halfsaturation or half-inhibition constants (Gafar et al., 2018). Different [H + ] tolerance ranges were detected at both the taxa level (e.g. between diatoms and coccolithophores) and on a species level (e.g. between coccolithophore species, Fig. 2). Crocosphaera watsonii (cyanobacteria), Coccolithus pelagicus, Gephyrocapsa oceanica and Scyphosphaera apsteinii (all three are large, heavily calcified coccolithophores and therefore particularly H + sensitive (Bach et al., 2015;Gafar et al., 2019)) were only able to grow up until c. 70 nmol l −1 H + (Table S2). By contrast, three diatoms species could grow at H + up to 170 nmol l −1 H + (Table S2). Hence, diatoms species considered in our analysis were considerably less sensitive to seawater pH F than many other phytoplankton taxa. No parameterization of the modified Monod equation was applied to Polarella glacialis (dinoflagellate) because of the highly variable response between the winter and summer populations. The winter population appeared to be considerably more sensitive to [H + ] than the summer population with a much smaller [H + ] window of maximum growth. Reaction norm variability between strains was also observed in Ceratium and Prymnesium (Fig. 2). The three Prymnesium parvum strains (haptophyte) and Chlorella vulgaris (chlorophyte) appeared to be robust up until at least 100 nmol l −1 H + (c. pH F = 7, Fig. 2), with peak growth rates in Chlorella vulgaris reported at 1000 nmol l −1 H + (i.e. pH F = 6, Mayo, 1997).

Universality of optimum curve reaction norm
Calcifying haptophytes are thought to be particularly sensitive to OA because of the formation of their calcium carbonate shells becomes increasingly energetically costly as seawater [H + ] increases (Monteiro et al., 2016). Here, highly taxonomically distinct phytoplankton taxa showed the same optimum curve reaction norm over a large range of H + concentrations. Sensitivity between phytoplankton groups with respect to inorganic carbon substrate and H + inhibitor was, however, highly variable (Fig. 2). Based on the evidence provided herein we argue that the optimum curve reaction norm will be found in every phytoplankton tested given that the applied pCO 2 range (under constant TA) is wide enough. Therefore, we suggest that the optimum curve response should be regarded as universal and be the underlying default reaction norm for phytoplankton growth under increasing CO 2 at constant TA.
Growth rates can be considered an integrative indicator of organism fitness and cell function (Boyd et al., 2013). Until a certain threshold, growth increases with increasing CO 2 (left side of the optimum) until the inhibitory behaviour presumably of H + becomes the dominant driver of the reaction norm (right side of the optimum). This response has been shown for coccolithophores and to some extent for diatoms and dinoflagellates (Bach et al., 2011;McMinn et al., 2014;Kottmeier et al., 2016;Shi et al., 2019) but it still needs confirmation in other species. Furthermore, the reaction norm is likely expandable to many other responses of physiological rates such as calcification (Langer et al., 2006;Bach et al., 2011Bach et al., , 2015Gafar et al., 2019), maybe even to other biota such as corals (Ries et al., 2010), as H + is an essential molecule for signalling and metabolic control in cells (Taylor et al., 2012). It is intriguing to think this response pattern in marine autotrophic organisms in general may be present for the same reason as coccolithophores.
The physiological explanations for this H + inhibition of growth may be similar between phytoplankton groups, even if the sensitivity to H + is different. For example, voltage gated proton (H v ) channels are protein complexes which can regulate cell pH homeostasis under rapid H + production in coccolithophores (Taylor et al., 2011) and their function is highly sensitive to the cross-membrane electrochemical, or H + concentration, gradient (DeCoursey, 2008). These channels are also present in the genomes of a wide variety of phytoplankton taxa including diatoms, dinoflagellates, chlorophytes (Taylor et al., 2012) although absent in prasinophytes (Taylor et al., 2011). Hence their relative ubiquity may explain the general H + growth inhibition response observed in diverse phytoplankton taxa. Nevertheless, cells may adopt different physiological strategies to maintain H + efflux across the plasmalemma. Two possible mechanisms are (1) increased energetic investment in active H + efflux to maintain cytosol pH or (2) adjustment of the cytosol pH or membrane potential to maintain the passive cross-membrane electrochemical gradient (Taylor et al., 2012). The latter has been identified in diatom species under a moderate pH decrease from 8.1 to 7.8 (Shi et al., 2019). These species (Phaeodactylum triconutum, Chaetoceros muelleri) also demonstrated the largest increase in growth to decreasing medium pH. This suggests that different strategies in maintaining H + efflux may also underlie species-specific differences in growth rates inhibition by H + . Particularly high [H + ] tolerance was observed for taxa within chlorophytes and noncalcifying haptophytes (here,  Prymnesium parvum) groups. Variable sensitivity within the dinoflagellate taxa to both low and high [H + ] is perhaps not surprising due to their diverse evolutionary history and trophic strategies (Taylor et al., 2009). Indeed, a link was recently established between carbon concentrating mechanism plasticity and the evolution of phytoplankton groups ( Van de Waal et al., 2019). While there is still insufficient data, we also expect such a link between the organism evolutionary histories or physiological traits and the reaction norm shape (Fig. 3). This hypothesis warrants further investigation.
Potential application of the optimum-curve response pattern to predict phytoplankton growth rates in the future ocean This literature compilation reveals variability in taxon-specific sensitivities that could be described in the simple empirical model (Fig. 2). Similar descriptions of taxon-related differences phytoplankton growth optima across large ranges in temperature (Eppley, 1972) and light (Goebel et al., 2008) have a long history and have had a large influence on planktological research. For example, temperature optimum curves from individual species form the basis of a generalized growth rate-temperature relationship ( 'Eppley Curve', Eppley, 1972), knowledge of considerable value to the modelling community (e.g. Thomas et al., 2012). Indeed, most scientists would probably agree that a complete temperature curve is more informative than comparing two temperature treatments, even though large parts of the response curve may be outside of the range organisms experience in their natural habitat. This underlines the value of our proposed experimental approach to explore physiological boundary conditions for phytoplankton groups in response to seawater pCO 2 , in contrast to selecting two or three CO 2 levels within the OA relevant range. We would expect taxon-specific ranges in CO 2 optima (Fig. 3) could be used to better understand shifts from one taxon to another in an acidifying ocean, as has been proposed for coccolithophores (Ridgwell et al., 2009). For example, using the approach of Dutkiewicz et al. (2015), this framework for phytoplankton taxa sensitivity could be integrated into ecological models as an extension of the Monod model currently employed.
Alternatively, the static model could be embedded in a dynamic energy budget to scale the response of individual phytoplankton physiology to population dynamics at an ecosystem level, similar to Muller (2011). We suggest this approach may be a particularly beneficial strategy for predicting the response of primary producers to projected changes in the marine environment. Studies investigating a CO 2 range well beyond what is relevant for OA could also be designed to probe interactive effects of CO 2 with other environmental factors. Addition of light, temperature, or macro/micronutrient treatments would reveal how these parameters modify the CO 2 response curve. While these environmental drivers may change the shape, the general optimum curve response pattern appears conserved (Sett et al., 2014;Bach et al., 2015;Gafar et al., 2019) and applicable to strains tested from different populations (Zhang et al., 2018).
The workload using this experimental approach due to the higher number of treatment levels required may be considered impractical. Yet, most studies in our literature analysis used three or four treatments in triplicate. Distributing the same number of Fig. 3 Conceptual figure of what multiple pCO 2 reaction norms for a variety of species/taxa could look like if there were sufficient datasets. Each line represents the reaction norm (optimum curve) of a different species or strain and shows intra-taxon variability, where some genotypes may be more or less sensitive to CO 2 . The reaction norms from the three taxa are shown here are used to demonstrate the optimum response curve concept, an Eppley curve-like equivalent for carbonate chemistry (Ridgwell et al., 2009), and are loosely based on results from the literature presented in Fig. 2.

Forum
New Phytologist bottles across individual treatment levels would give nine to 12 distinct treatments, which would have been much more informative with minimal difference in work load. Hence, here we hope to provide an impetus to reignite enthusiasm for this mechanistic approach to characterize phytoplankton reaction norms to CO 2 . By increasing the number of treatments, here CO 2 , at the expense of replication enables quantification of the response as well as allowing for statistically robust description of nonlinear response patterns (Cottingham et al., 2005) thereby deepening our understanding of carbonate chemistry control of phytoplankton physiology.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.    Methods S1 Culture conditions and growth rate determination for Crocosphaera watsonii.
Methods S2 Carbonate chemistry (total alkalinity, dissolved inorganic carbon concentration) Notes S1 Evidence for CO 2 as driver of growth rates at low pCO 2 .
Table S1 Summary of final and initial cell densities, total incubation time, calculated growth rate and pH in Crocosphaera watsonii culture experiment.

Table S3
List of studies and corresponding experiment design details found in the literature reporting on the response of phytoplankton taxa to CO 2 /pH. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.