Carbonic anhydrases are influenced by the size and symbiont identity of the aggregating sea anemone Anthopleura elegantissima

Many cnidarians engage in symbioses with photosynthetic microalgae. These holobionts form the structural and trophic bases for coral reef ecosystems and are major components of temperate intertidal ecosystems (reviewed in Shick, 1991). The importance of these symbioses to these ecosystems lies in the nutritional exchange between the host and microalgal symbionts (Harland and Davies, 1995; Muscatine, 1990). The temperate symbiotic sea anemone Anthopleura elegantissima is a prominent member of the intertidal ecosystem on the west coast of North America. These anemones rely primarily on heterotrophic nutrition but are facultatively symbiotic and occur in three symbiotic states depending on light intensity and temperature: brown, green and aposymbiotic (Saunders and Muller-Parker, 1997). Brown anemones obtain their color from the presence of the dinoflagellate symbiont Breviolum muscatinei (formerly Symbiodinium, ITS2 type B4; Lajeunesse and Trench, 2000; LaJeunesse et al., 2018) and occur in high light, warmer environments. In contrast, green anemones harbor the chlorophyte Elliptochloris marina and are found in low light, cooler habitats (Letsch et al., 2009; Lewis and Muller-Parker, 2004). Aposymbiotic anemones are white, symbiont-free and are found in very low light or in darkness (Secord and Augustine, 2000). Although both species of microalgae provide photosynthate to their host anemone, B. muscatinei is more productive than E. marina, thus requiring more dissolved inorganic carbon (DIC) for photosynthesis and passing more organic carbon to the host (Verde and McCloskey, 2007).

Carbonic anhydrases (CAs; EC 4.2.1.1) are critical for DIC transport in symbiotic cnidarians (Al-Moghrabi et al., 1996; Furla et al., 2000a,b; Le Goff et al., 2016; Weis, 1991; Weis and Reynolds, 1999; Weis et al., 1989). Microalgal symbionts are housed deep in host tissue within host gastrodermal cells, making the diffusion of DIC and oxygen from the surrounding seawater difficult. This challenge is compounded by symbiont productivity and animal size because as symbiont productivity increases, so does the requirement for DIC, and as animal size increases, surface-area-to-volume (SA:V) ratios decrease, providing relatively less area for diffusion across cell membranes. Consequently, as expected, Weis et al. (1989) demonstrated that SA:V ratio also affects CA activity when compared across symbiotic cnidarian species.

Previous studies have shown that CA activity is higher in cnidarians with symbionts compared with those without (aposymbiotic; Weis, 1991; Weis and Reynolds, 1999; Weis et al., 1989). Weis and Reynolds (1999) measured gene expression of a cytosolic CA in A. elegantissima using semi-quantitative PCR and found that brown A. elegantissima had higher expression than aposymbiotic anemones, matching the differential enzyme activity. These studies suggest that CA is inducible and depends on the presence of symbiotic algae.

Previous work suggests that A. elegantissima acquires most of its daily nutrition via heterotrophy (Bergschneider and Muller-Parker, 2008). This sets the stage for updating our understanding of how CAs, enzymes that function in autotrophy in symbiotic cnidarians, participate in the trophic strategy of a symbiotic organism under changing environmental conditions (Fig. 1). In this study, we investigate CAs in A. elegantissima in more detail by identifying CA paralogs and their phylogenetic placement, thereby inferring putative function and location in the cell. We extend the examination of the effect of symbiotic state on CA activity and gene expression of multiple CA paralogs, and provide insight into the effects of animal size and light environment on activity. Consequently, we wanted to answer the following questions: (1) what are the relative concentrations of CA as a function of symbiotic state; and (2) how do a physical factor (light intensity) and a biological factor (anemone size) affect CA activity?

Experimental manipulations and CA activity measurements of the symbiotic state study were performed at Walla Walla University's Rosario Beach Marine Laboratory (RBML), Anacortes, WA, USA. The light and anemone size experiments on CA activity measurements, and gene expression analyses of the symbiotic state experiment, were conducted at Oregon State University (OSU).

The present study used adult Anthopleura elegantissima (Brandt 1835) collected at low tide during the summer of 2017 from Sares Head, Fidalgo Island, WA, USA (symbiotic and size experiments), or Lawrence Point, Orcas Island, WA, USA (light experiment). Anemones were removed from rocks using a microspatula, cleaned of all foreign materials (shell and macroalgal bits), immediately flash frozen in liquid nitrogen (symbiotic state and size experiments) or transported back to RBML (light experiment), placed into labeled 118 ml glass jars, and maintained in an outdoor flow-through seawater tank with nylon-stocking-filtered seawater. Flash-frozen anemones were subsequently stored at −80°C until processed. The symbiotic state of anemones in the field was determined by eye during collection and subsequently confirmed in the laboratory by examining tentacle squashes under a microscope. Brown and green anemones were classified as having at least 90% of their respective symbionts, though there were very few anemones that had mixed algal populations.

For the symbiotic state experiment, aposymbiotic anemones (n=10), anemones containing Elliptochloris marina (green anemones, n=10) and anemones containing Breviolum muscatinei (brown anemones, n=10) were collected from three distinct clones within 3 m of each other. Clones are easily distinguishable by the presence of anemone-free zones that typically separate clonal aggregations and the absence of directed, aggressive acrorhagi deployment (Ayre and Grosberg, 2005; Francis, 1973). When emerged at low tide, A. elegantissima retract their tentacles, thereby preventing use of tentacle crown or oral disk diameter as a size measure. Instead, we measured the column crown diameter of each anemone and used it as a proxy for total anemone size. For the size experiment, 20 brown anemones from each size class [small (0–14 mm); medium (15–29 mm); large (>30 mm)] were collected from one clone. For the light experiment, 30 each of aposymbiotic, green and brown anemones were collected.

Sea anemones were collected under Washington Fish and Wildlife permit NESTLER 17-103 to Jim Nestler (Rosario Beach Marine Laboratory, Walla Walla University).

An outdoor flow-through seawater tank was modified into three compartments (from inflow to outflow): a cover with several layers of black plastic sheet was added to create a dark environment, neutral density screens were added for a medium light environment, and the last compartment was unmodified for full exposure to the sunlight (high light environment). Light experiment anemones of each symbiotic state (n=10) were randomly divided into the three treatment groups: high (1400 µmol photons m⁻² s⁻¹), medium (250 µmol photons m⁻² s⁻¹) or no light. Light intensity in each compartment was measured around 11:00 h daily using the sunlight setting on a QMSW-SS Quantum light meter (Apogee, Logan, UT, USA) and are reported as 28-day averages of single point measurements.

Twenty-four hours after collecting the anemones, they were placed into their respective light treatments and allowed to acclimate for 7 days. Thereafter, five anemones of each symbiotic state and light treatment were flash frozen in liquid nitrogen as a baseline (day 0) for CA activity. Twenty-eight days after the initial sampling, the remaining anemones were flash frozen (day 28) and analyzed for changes in CA activity from the baseline measurements. Every 2 days during acclimation and treatment, anemones were rotated within their treatment areas to eliminate any light differences within compartments. Jars were changed and cleaned every 7 days of the experiment. Over the course of the experiment, the temperature of the seawater ranged from 12.7 to 13.7°C.

To quantify algae in hosts, a small piece of tissue was cut from each host anemone to include tentacles, oral disc and column tissue. The anemone tissue was homogenized in a known amount of filtered seawater (FSW) in a 2 ml ground-glass homogenizer until no large chunks remained. The resulting homogenate was vortexed for 10 s and algal density was counted in triplicate using a Countess II automated cell counter (Invitrogen, Singapore). To index algal number to host protein, remaining homogenates were centrifuged at 4200 g in a tabletop centrifuge for 10 min at 4°C to pellet the algae and other cellular debris. The anemone-only protein supernatant was decanted and soluble protein concentration was determined spectrophotometrically using a Bradford protein assay (Bradford, 1976) following the manufacturer's instructions. Triplicate algal counts were averaged and normalized to soluble anemone protein.

CA activity was measured as previously described (Weis and Reynolds, 1999) (for full protocol details, see dx.doi.org/10.17504/protocols.io.j4hcqt6). Briefly, on ice in a glass-Teflon homogenizer, one-quarter of each anemone was homogenized in cold enzyme extraction buffer. The resulting homogenate was transferred to a 15 ml tube and centrifuged at 4200 g for 10 min at 4°C to pellet the algae and cellular debris. The anemone supernatant was decanted and divided into three aliquots destined for native (3 ml), denatured (3 ml) and protein (0.5 to 1 ml) measurements. The denatured sample was boiled for 5 min and placed on ice. In a 6 ml flat-bottom test tube equipped with a tiny stir bar, 1 ml of anemone fraction and 1 ml of cold Buffer A were combined while on ice. When the pH stabilized at 8.2, 1 ml of CO₂-saturated cold deionized water was quickly added. The change in pH of this mixture was recorded using a PH47-SS stainless steel pH probe (Hach, Loveland, CO, USA) attached to an H260GNP pH meter (Hach) writing data to a computer running H-series SmartLogger II (Hach) until the pH stabilized. This was repeated a total of three times for the native and denatured samples. Anemone protein concentration was determined spectrophotometrically using a Bradford protein assay (Bradford, 1976).

Phylogenetic analysis aids in predicting the subcellular location of proteins based on their sequences, domain architecture and groupings with known, well-researched orthologs from other species. Known and predicted AeleCA gene sequences were obtained from Weis and Reynolds (1999) (accession AF14053) and by a gene name search for ‘carbonic anhydrase’ in the aposymbiotic A. elegantissima transcriptome v1.0 database. The working assembly transcriptome for A. elegantissma is accessible on Dryad (datadryad.org/resource/doi:10.5061/dryad.3f08f) and raw data are available on NCBI (SRA accession: SRR2300240) (Kitchen et al., 2015a,b). The nucleotide sequences specific for AeleCAs examined in this study are provided in Table S1. Published cnidarian, sponge and human CA protein sequences were obtained from appendix S1 and table S1 in Le Goff et al. (2016).

Open reading frame (ORF) analysis and translation, in all six frames, was performed on all potential AeleCA sequences using Geneious (v10.2.6). Homology of AeleCAs was assessed using Blastp against the NCBI non-redundant database and domain analysis with Pfam. AeleCAs with top hits to other CAs were retained for further domain architecture analyses. Signal peptide, transmembrane domains and subcellular localization motifs were predicted using SignalP (v4.0), TMHMM (v2.0) and TargetP (v1.1), respectively (Table S1). From these analyses, we picked four AeleCA genes representing different predicted subcellular localizations. Protein sequences specific for AeleCAs examined in this study are provided in Table S1.

Protein sequences from Le Goff et al. (2016) and AeleCAs were imported as a single matrix in Mesquite (v3.6) with Zephyr (v3.0) for phylogenetic analyses (http://www.mesquiteproject.org and http://zephyr.mesquiteproject.org). To map the predicted subcellular localization from the Le Goff et al. (2016) phylogenetic tree to ours, we created a taxon matrix with the taxa and predicted CA localization for each. AeleCAs were trimmed of signal peptides and terminal transmembrane domains to only include the section corresponding to the CA catalytic domain as performed in Le Goff et al. (2016). Sequences were aligned using Muscle (v3.8.31) and both model selection and tree inferences were performed using IQTREE ModelFinder and IQTREE nonparametric bootstrap approximation, respectively (Edgar, 2004; Kalyaanamoorthy et al., 2017). The Mesquite and IQTREE files are available at https://github.com/jackckoch/aele.ca.

Preliminary tree inferences suggested that there were several potential rogue taxa in our analysis. Rogue taxa can create instability and generate poor support in phylogenetic analyses, which can obscure and reduce support for relationships among more stable taxa. To identify rogue taxa in our dataset, we used RogueNaRok with an IQTREE nonparametric bootstrap approximation with 100 replicates (Aberer et al., 2013; Nguyen et al., 2015). RogueNaRok identified several potential rogue taxa with one being the most outstanding: Hydra magnapapillata CA13. Removal of this taxon improved the relative bipartition information criterion (RBIC) of our tree from 0.502 to 0.529. The RBIC is the sum of all support values divided by the maximum possible support in a fully bifurcating tree, thus a higher RBIC value indicates a larger percentage of the total support possible (Aberer et al., 2013). In our preliminary tree iterations, H. magnapapillata CA13 jumped around in the tree, often appearing as the longest branch in the tree; therefore, based on the rogue taxa analysis, this taxon was removed from future tree iterations. Following rogue taxa removal, we re-ran RogueNaRok to check for other potential rogue taxa, but found none that were as outstanding or concerning as H. magnapapillata CA13.

According to IQTREE ModelFinder, the best model was WAG+F+R6 according to the Bayesian information criterion (BIC; −Lnl=42,453.063) and WAG+F+R7 according to Akaike's information criterion (AIC; −Lnl=42,450.721), which are both similar to the model used in Le Goff et al. (2016). These models utilize general amino acid exchange rate matrices (WAG; Whelan and Goldman, 2001) that have empirical amino acid frequencies from the data (+F), and a FreeRate model that generalizes the discrete gamma model by relaxing the assumptions of the gamma-distributed rates (+R; Soubrier et al., 2012; Yang, 1995) by the number of categories specified after the model modifier (e.g. +R6 or +R7). AIC and BIC are sets of criteria for model selection. In our analysis, tree building used the stricter, BIC best model and we used IQTREE to create a maximum likelihood tree and evaluated 1000 nonparametric bootstrap approximation replicates to assess branch support (Fig. 2) (Nguyen et al., 2015).

Total RNA was extracted from anemones collected for the symbiotic state comparison (n=5 per symbiotic state) using a hybrid extraction protocol with TRIzol Reagent (Life Technologies, Eugene, OR, USA) and an RNeasy Mini kit (Qiagen, Valencia, CA, USA) as described by Poole et al. (2016) and stored at −80°C. We performed a cleanup step on all samples following the protocol as described in Poole et al. (2016). Both RNA quality and quantity were assessed using a NanoDrop ND-1000 (NanoDrop, Wilmington, DE, USA) and a Qubit 3.0 fluorometer (Invitrogen, Carlsbad, CA, USA), respectively. The cDNA synthesis was performed with the Protoscript II first strand synthesis system (New England Biolabs, Ipswich, MA, USA) using 150 ng of RNA and the Oligo(dT) primer.

Prior to performing qPCR on the AeleCA genes, reference genes that had stable expression across symbiotic states were determined. We identified several common reference genes used in qPCR experiments with cnidarians [DNA-directed RNA polymerase, 60S ribosomal protein L11, glyceraldehyde 3-phosphate dehydrogenase (G3PD), 40S ribosomal protein S7, and nicotinamide adenine dinucleotide (NADH) dehydrogenase; Kenkel et al., 2011; Kitchen et al., 2015b; Leggat et al., 2011; Lehnert et al., 2014; Poole et al., 2016; Radonić et al., 2004; Seneca et al., 2010] and evaluated their expression levels from A. elegantissima symbiotic state RNA-seq data (J. C. Koch, personal observation). Of the genes identified, NADH dehydrogenase and G3PD had the most stable expression across the three anemone symbiotic states.

Specific primers for each AeleCA were designed using Primer3 in Geneious (v10.2.6) to produce 149 bp amplicons (Table S2). Primer efficiencies and optimal template concentration were determined using standard curve analysis with a two-fold dilution series of cDNA from a brown anemone sample and were between 87 and 100% (https://github.com/jackckoch/aele.ca). For each biological replicate, qPCR was performed in triplicate using the Power SYBR^® Green PCR Master Mix (Applied Biosystems, Waltham, MA, USA) with premixed ROX in a Mastercycler Realplex4 ep Gradient S (Eppendorf, Germany) under the standard settings, using the following run protocol with the addition of a melt curve: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 59°C for 45 s, and 72°C for 15 s. We followed the manufacturer's instructions with the following modification: the total reaction volume was 15 µl, which included 0.45 µl of each forward and reverse primer at 10 µmol, 2 µl of the appropriate cDNA template (total concentration per reaction of 15 ng µl⁻¹), 7.5 µl of Power SYBR Green PCR Master Mix, and 4.6 µl of nuclease-free water. Plates included no-template, no-reverse transcriptase and no-primer controls, and at least one interplate calibrator (IPC).

Data were analyzed in RStudio (v1.1.463) (https://www.rstudio.com/) with R (v3.4.2) (https://www.R-project.org/) and figures were generated with the ggplot2 (v2.2.1) package (Wickham, 2009). R code is available at https://github.com/jackckoch/aele.ca.

Normality of distribution was tested with the Shapiro–Wilk test (P<0.05) and by visually assessing the distribution of the residuals in the quantile–quantile plots. Data were not normally distributed; however, ANOVAs are robust against normality violations so we ran both parametric and non-parametric analyses. Homogeneity of variance was tested with the Levene's test (P>0.05). Differences in CA activity between symbiotic states were tested using a Kruskal–Wallis test (P<0.05) followed by a Benjamini–Hochberg corrected post hoc Dunn's test (non-parametric) and a two-way ANOVA (P<0.05) followed by a post hoc Tukey's honest significant difference (HSD) test (parametric). Both tests provided the same results. Differences in CA activity between light environment start and end values were tested using a Student's or Welch's t-test (P<0.05) where appropriate. Linear models were used to determine which independent variables (e.g. diameter, algal density) best predicted a dependent variable (CA activity).

For CA gene expression, all plates had their baselines and thresholds set to 3–15 and 800, respectively. The crossing threshold (C_t) values from the IPCs across all plates were averaged and necessary corrections for each plate were calculated from that average. Corrected C_t values for qPCR replicates were averaged and imported into RStudio for analysis using the Markov chain Monte Carlo (MCMC) qPCR package (Matz et al., 2013), which converts the C_t values to transcript counts, allowing the analysis of relative gene expression instead of absolute expression. The counts were grouped by gene and symbiotic state and plotted to evaluate the abundance of each gene. We used 39 as the C_t value that represents a single transcript and although this number was not verified on the qPCR machine, we treated all samples the same, lending validity to our analysis of relative gene expression. Aposymbiotic anemones were set as the control group in the experiment. First, we fit a naïve model without reference genes. The summary plot from this model suggested that NADH dehydrogenase and G3PD both had some variability in expression amongst our treatment groups that needed to be accounted for in our informed model. We fit an informed model (which included normalization to the reference genes) and took into account variation in the stability of reference genes (m.fix=1.5). Differences in the resulting relative quantities on the log₂ scale were tested using Student's t-tests (P<0.05). Although the model took into account the control group (aposymbiotic anemones) variance, we estimated and visualized the aposymbiotic group variance by setting one of the experimental groups (brown or green anemones) as the control and rerunning our model. When green anemones were set as the control, the credible intervals for the aposymbiotic anemones were larger than when brown anemones were set as the control. For this reason, we modeled green anemones as the control in order to illustrate the credible intervals for the aposymbiotic anemones.

We initially identified many potential AeleCAs (https://github.com/jackckoch/aele.ca), but selected four CA paralogs with different putative subcellular locations – secreted, secreted or membrane-bound, membrane-bound, and cytosolic – based on previous research, prediction programs and by adding our AeleCAs to a cnidarian CA phylogeny from Le Goff et al. (2016) (see Materials and Methods, Fig. 2). We named these genes by human CA convention: cytosolic, AeleCA1; secreted, AeleCA6; membrane-bound, AeleCA4; secreted or membrane bound, AeleCA9. The resulting phylogenetic tree shows that CAs form distinct clusters based on putative subcellular localization, albeit with weaker support than other published trees (Fig. 2). Most sequences grouped into phylum-specific clusters with AeleCAs always grouping with symbiotic sea anemone Exaiptasia pallida CAs within cnidarian clades.

Aposymbiotic and green anemone CA had similar average activities of 0.28±0.09 and 0.38±0.13 −ΔpH min⁻¹ mg⁻¹ soluble animal protein, respectively (n=10), whereas brown anemones had significantly higher average activity of 0.79±0.11 −ΔpH min⁻¹ mg⁻¹ soluble animal protein (ANOVA, F=65.65, P<0.001; and Kruskal–Wallis test, χ²=20.612, P<0.001; n=10; Fig. 3).

For cytosolic AeleCA1, there were no significant differences in expression between the three anemone types (Student's t-tests: brown–green P_MCMC=0.30, brown–aposymbiotic P_MCMC=0.81, green–aposymbiotic P_MCMC=0.43; Fig. 4A). For putative membrane-bound AeleCA4, brown and green anemones had 0.724 and −0.072 mean relative expression, respectively. Brown anemone expression was significantly higher than green anemone expression, but neither were statistically different from that of aposymbiotic anemones (Student's t-tests: brown–green P_MCMC<0.05, brown–aposymbiotic P_MCMC=0.06, green–aposymbiotic P_MCMC=0.80; Fig. 4A). For putative secreted AeleCA6, brown and green anemones had 1.113 and 0.256 mean relative expression, respectively. Brown anemone expression was significantly higher than aposymbiotic anemone expression (Student's t-test: brown–aposymbiotic P_MCMC<0.05). Brown anemone expression compared with green anemone expression and green anemone expression compared with aposymbiotic anemone expression were not statistically different from each other (Student's t-tests: brown–green P_MCMC=0.10, green–aposymbiotic P_MCMC=0.69; Fig. 4A). For putative membrane-bound or secreted AeleCA9, brown and green anemones had −2.596 and 1.542 mean relative expression, respectively, which were significantly different from each other and from aposymbiotic anemone expression (Student's t-tests: P_MCMC<0.001; Fig. 4A).

We compared the transcript abundance with relative expression, allowing us to illustrate the putative importance of each paralog for each symbiotic state (Fig. 4B). AeleCA1 had the highest transcript abundance for all three symbiotic states, followed by AeleCA6 and AeleCA4 (Fig. 4B and Fig. S2). For all three paralogs, brown anemones had the highest mean abundance. In contrast, AeleCA9 was most abundant in green anemones, followed by aposymbiotic and brown. AeleCA9 in brown anemones was the least abundant overall.

CA activity in brown anemone tissue showed a significant inverse correlation with the size of the anemone (linear regression: adjusted R²=0.5424, P<0.001; Fig. 5).

With the exception of green anemones exhibiting a significant decrease in average activity after 28 days in the dark compared with day 0 (Student's t-test: P<0.05; Fig. 6), all other anemone treatment groups showed no significant change in average CA activity after 28 days in their respective light environments with all data points present (Student's t-test: P>0.05; Fig. 6). However, the day 28 anemones in the green high light and brown darkness groups each had an outlier present as indicated in Fig. 6. After removing the outlier from the day 28 green anemone high light group, there was no significant difference compared with day 0. In contrast, after removing the outlier from the day 28 brown anemone darkness group, there was a significant difference compared with day 0 (Welch's t-test: P<0.001; Fig. 6).

Regardless of symbiotic state, there were no significant correlations between algal density and CA activity (Figs S1–S3).

Green anemone CA activity was surprisingly lower than expected given that these anemones have a photosymbiont that requires CO₂. The low CA activity in green anemones (not different than aposymbiotic animals; Fig. 3) suggests that green algae DIC demand is met or exceeded by host respiration and diffusion from seawater. Bergschneider and Muller-Parker (2008) came to a similar conclusion from their work examining stable isotope signatures of different A. elegantissima symbiotic states. Other research has shown that photosynthetic rates in E. marina are much lower than in B. muscatinei (Verde and McCloskey, 2002), further supporting the low CA activity measured in our study. Decreased demand for DIC in green anemones could also be explained by differences in the carbon fixing enzyme ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO) between the two algal species. Dinoflagellates contain RubisCO form II whereas green algae have form I (Morse et al., 1995; Tabita et al., 2008); form I is more efficient, with a higher affinity for CO₂. Therefore, E. marina may be able to acquire enough CO₂ to satisfy DIC needs through diffusion alone whereas B. muscatinei cannot, thus inducing the host to increase CA synthesis and activity as a compensatory mechanism.

Our observed high CA activity in brown versus aposymbiotic anemones recapitulates earlier experiments on A. elegantissima (Weis and Reynolds, 1999); however, expression patterns between the two studies are not congruent. We showed no difference in AeleCA1 (cytosolic) gene expression between symbiotic states (Fig. 4A), a pattern opposite of that observed by Weis and Reynolds (1999). AeleCA4 (putative secreted or membrane-bound) and AeleCA6 (putative secreted) expression patterns matched the overall CA activity patterns. AeleCA1 and likely AeleCA6 are internal CAs (cytosolic and not attached to a cell membrane) (Weis and Reynolds, 1999). It is more difficult to determine whether AeleCA4 and AeleCA9 are internal or external because they could localize to internal or external membranes (Fig. 7). Further research to definitively localize A. elegantissima CAs will be critical to fully interpret the results of our study in the context of the symbiotic relationship.

The CA activity assay does not discriminate between different CA paralogs and it is likely that not all CA protein types were captured in animal homogenizations. Extracellular CA might be lost and any membrane-bound CAs might be pelleted with membranes and therefore not end up being measured in the assay. Furthermore, we made no measurement of CA protein turnover, so expression quantity and protein quantity are not necessarily congruent.

The observed pattern in AeleCA9 (putative secreted or membrane-bound) of high abundance and expression in green anemones and low in brown compared with aposymbiotic anemones suggests that this CA is not an important player in brown anemones (Fig. 4B and Fig. S2). The high expression and abundance of AeleCA9 in green anemones could indicate that the presence of E. marina induces different DIC pathways than does the presence of B. muscatinei (Fig. 7). These pathway changes could include differential expression and use of CAs, H⁺-ATPases, HCO₃⁻ transporters and paracellular pathways (Allemand et al., 1998; Barott et al., 2015a,b; Furla et al., 2000a; Oakley et al., 2016; Tresguerres et al., 2017).

The patterns displayed by AeleCA9 have parallels to the results of Tansik et al. (2015, 2017), who used membrane inlet mass spectrometry (MIMS) to measure CA activity in corals and their symbionts by calculating the ¹⁸O rate of exchange and following labeled ¹³CO₂ species. They parsed internal (cytosolic) and external (facing outwards or secreted outside of the cell membrane) CA activity measurements by blocking external CAs with the membrane-impermeable CA inhibitor acetazolamide. Tansik et al. (2015, 2017) showed that algal symbionts of the corals Orbicella faveolata, Porites astreoides and Siderastrea radians had internal CA activity values that correlated with their maximal photosynthesis rate rank: P. astreoides, S. radians and O. faveolata (highest to lowest average rate) (Tansik et al., 2017). Porites astreoides and S. radians had the lowest external CA activity, yet their symbionts were the most productive. These two corals mirror the pattern of putative external AeleCA9 expression and symbiont productivity in brown A. elegantissima in that B. muscatinei have higher productivity compared with E. marina (Tansik et al., 2017; Verde and McCloskey, 1996). This suggests that the Symbiodiniaceae species of all three corals and A. elegantissima are capable of maximal photosynthesis with limited DIC (Tansik et al., 2017) and could indicate a host-controlled carbon environment as proposed by Davy et al. (2012) and Weis (1993) to limit symbiont growth, reproduction and photosynthetic rates in favor of passing photosynthate to their cnidarian hosts.

In contrast, O. faveolata had the highest external CA activity, providing ample DIC for its low-productivity symbionts. This pattern mirrors what we observed in green anemones, which had high AeleCA9 gene expression (Fig. 4A) and low photosynthetic rates compared with their brown counterparts. We propose that green anemones have adequate external CA activity from AeleCA9 that, in turn, provides a carbon-sufficient environment for their symbionts (Fig. 7), but that AeleCA9 activity was not captured by our activity assay because AeleCA9 proteins may be pelleted with the host membranes (see Materials and Methods, and above).

The congruence between photosynthetic rate and host DIC supply in Tansik et al. (2017) suggests that the corals in their study have adapted CAs to support their algal symbionts. In contrast, our work with A. elegantissima suggests that CAs are influenced by the phenotype of the symbiosis and that CAs could change if symbiont identity changes (see Bates, 2000 for symbiont switching in Anthopleura). Future studies using a single clone in different symbiotic states to examine the subcellular location for the different CA paralogs and tissue-specific CA activity and expression could help reveal their roles in the symbiosis.

Anemone size had a strong, inverse effect on CA activity in brown A. elegantissima of differing column crown diameters (Fig. 5). Our findings contrast previous work in symbiotic cnidarians that found cnidarians with low surface-area-to-volume ratios to have high CA activity compared with cnidarians with higher surface-area-to-volume ratios (Weis et al., 1989). Of note, size-based CA activity in our study is within or above the range of activity measured for brown anemones in the symbiotic state experiment.

The inverse correlation between column crown diameter and CA activity might be explained by a combination of trophic strategy and buffering potential that would determine how much CA is needed to satisfy algal DIC demands and maintain cellular homeostasis. Using stable isotope analysis, previous research has shown that A. elegantissima generally relies more heavily on heterotrophy to satisfy daily nutritional needs rather than autotrophic carbon supplied by their algal symbionts (Bergschneider and Muller-Parker, 2008). Larger individuals could therefore be acquiring a greater percentage of their daily carbon and nitrogen nutrition by heterotrophic feeding while smaller individuals could be relying more heavily on symbiont autotrophy. This idea is supported by previous research showing that larger individuals have greater tentacle and prey capture surface area (Sebens, 1981; Zamer, 1986) and research using stable isotope analysis showing that larger individuals from the genus Anthopleura rely more heavily on heterotrophy to satisfy daily nutritional needs compared with smaller anemones (Bedgood and Bracken, 2020). Thus, there may be a greater demand for DIC in smaller anemones from their algal symbionts, thereby inducing higher CA activity. Furthermore, algae symbiotic with smaller A. elegantissima are fixing more carbon per day relative to large A. elegantissima (E. A. Verde, personal observation), also supporting the increase in CA activity that we measured in smaller individuals.

Smaller body sizes favor increased diffusion of gases (e.g. CO₂ and O₂) related to an increased surface-area-to-volume ratio (Brown and Sibly, 2006; Pörtner, 2002), and this may reduce the buffering capacity of small organisms to environmental change. In addition to its role in symbiosis, CA also plays a role in intracellular pH homeostasis (Tashian, 1989). Small anemones may be increasing CA activity to increase their capacity to maintain pH homeostasis. Consequently, the high CA activity observed in small anemones might be induced by a combination of trophic and buffering capacity differences with anemone size.

Given that algal productivity is dependent on light level (Dimond et al., 2013; Verde and McCloskey, 2002, 2007), we predicted that light intensity would affect carbon-concentrating mechanisms for DIC delivery to algae in A. elegantissima, as measured by CA activity. We hypothesized that activity in brown and green anemones in the dark, and green anemones in high light would decrease the most compared with brown and green anemones in high and medium light, respectively, because photosynthesis would cease or be affected by photoinhibition, thereby reducing the need for CA-supplied DIC. Our results for the two symbiotic groups placed in darkness supports this hypothesis and conclusions from research by Dimond et al. (2013) that examined symbiont photo-physiology under different light intensities. Green anemones placed in high light for 28 days showed a modest decrease in CA activity after removing an outlier, but the change was not significant. For our control groups (i.e. groups that were kept in their respective light environment) and the other light groups, there were no significant changes in CA activity with light.

There are several possible explanations for this stability in activity in some groups across light levels. Firstly, the anemones could have acclimated to their experimental light environments during the 7-day acclimation period of the experiment (see Materials and Methods). CA half-life and degradation rates were not examined in our study but are known to be variable between different tissue types and different organisms (30–80 h in different mouse tissues: McAvoy et al., 1981; 8 h half-life in Chlamydomonas: Toguri et al., 1989). Secondly, the experiment duration could have been too short, although with an experimental exposure to the various light regimes for 28 days, this possibility seems remote for most groups. A possible exception may be for the green anemones in high light that, given more time, might have decreased CA activity even further as the effects of high-light-induced algal photoinhibition reached the anemone. Thirdly, we did not control for animal size in this experiment. Based on our result of differential CA activity with animal size (see Fig. 5), it is possible that size differences masked a light effect. Finally, light simply may not play a large role in CA induction because photosynthetic symbionts are not the primary suppliers of host nutrition. Anthopleura elegantissima may instead rely largely on heterotrophic nutrition (Bergschneider and Muller-Parker, 2008; Levine and Muller-Parker, 2012) and might therefore be less responsive to changes in light, especially in the short term. Instead of a change in autotrophy-related nutrition (e.g. a change in CA activity), Hiebert and Bingham (2012) showed a general increase in heterotrophy by brown, green and aposymbiotic anemones in sunlight compared with in shade, supporting our findings and a general dominance of heterotrophy in A. elegantissima regardless of symbiotic state. However, seasonal changes in CA activity are a strong possibility because light regimes, temperature profiles and zooplankton density are lower during the winter than during summer months (Dimond et al., 2013; Verde and McCloskey, 2007).

Our study examined the effects of symbiosis, animal body size and light on CAs in a temperate symbiotic sea anemone and their integration into animal trophic strategy (Fig. 1). We found that not just one, but all three factors contribute to the dominance of heterotrophy in A. elegantissima. Our data suggest that, based on the behavior of CAs, symbiosis with autotrophic algae is a source of nutrition for these anemones, more so in brown anemones compared with green and aposymbiotic anemones, but as an accessory to heterotrophy in these organisms.

We thank Brian Bingham, Jay Dimond and Nate Schwarck at the Shannon Point Marine Center for their advice and assistance in locating anemones, and Nancy Nestler, Jim Nestler, Chris Lindsey, Flora Gibbs, Katie Pekar and Monica Culler at the Rosario Beach Marine Laboratory for their assistance in collecting anemones. Many thanks to Billie Swalla at Friday Harbor Laboratories for use of a plate reader for protein analyses. Thank you to David Maddison and Antonio Gomez for advice and assistance with our phylogenetic analysis and Mesquite usage, and Doug Warrick and Mike Brawner for their discussions about metabolism and scaling. Thanks to Trevor Tivey, John Parkinson, Katherine Dziedzic and Eli Meyer for their assistance with qPCR and advice on data analysis. Special thanks to Valeri Lapacek, Jason Presnell, Allie Graham, Shumpei Maruyama, Nathan Kirk and two reviewers for their discussions and critical comments, which helped to improve the manuscript.