Seagrass can mitigate negative ocean acidification effects on calcifying algae

The ultimate effect that ocean acidification (OA) and warming will have on the physiology of calcifying algae is still largely uncertain. Responses depend on the complex interactions between seawater chemistry, global/local stressors and species-specific physiologies. There is a significant gap regarding the effect that metabolic interactions between coexisting species may have on local seawater chemistry and the concurrent effect of OA. Here, we manipulated CO2 and temperature to evaluate the physiological responses of two common photoautotrophs from shallow tropical marine coastal ecosystems in Brazil: the calcifying alga Halimeda cuneata, and the seagrass Halodule wrightii. We tested whether or not seagrass presence can influence the calcification rate of a widespread and abundant species of Halimeda under OA and warming. Our results demonstrate that under elevated CO2, the high photosynthetic rates of H. wrightii contribute to raise H. cuneata calcification more than two-fold and thus we suggest that H. cuneata populations coexisting with H. wrightii may have a higher resilience to OA conditions. This conclusion supports the more general hypothesis that, in coastal and shallow reef environments, the metabolic interactions between calcifying and non-calcifying organisms are instrumental in providing refuge against OA effects and increasing the resilience of the more OA-susceptible species.

Seagrass meadows and calcifying algae beds are benthic communities that play unique roles in the removal, storage and release of carbon from seawater, via photosynthesis and/or calcification 1 . Coastal communities are metabolically responsible for 85% of the organic carbon and 45% of the inorganic carbon (C i ) buried in coastal sediments [2][3][4] . CO 2 is essential to photosynthesis, yet its increase in seawater reduces pH and carbonate ions, threatening the calcification process 5 . However, these ecosystems naturally experience large vertical and horizontal variations in abiotic parameters, namely pCO 2 and temperature 6 , that can vary from 400 to 10,000 µatm 2 and 15 to 30 °C 7 , respectively. Research has suggested that exposure to natural fluctuations alongside possession of phenotypic plasticity may help organisms and populations to resist or acclimate to novel anthropogenic conditions 6,8 .
Little is known about the existing interactions between calcifying and non-calcifying primary producers under OA and temperature rise. Whether it is via alteration of seawater chemistry, allelopathy or other molecular signaling, neighboring marine plants interact by influencing each other's metabolisms 9 . Changes in benthic macrophyte communities are projected for the future 10 where altered competition dynamics between fleshy and calcifying algae already have been shown to drive ecosystem shifts under elevated CO 2 conditions 11 . The current incomplete understanding of these interactions and the consequent mechanisms that drive ecosystem changes limit our ability to make realistic predictions for the effects of OA and warming on future community structure.
Seagrasses can act as buffers to OA by absorbing large quantities of CO 2 and increasing the pH of seawater [12][13][14] . Diel pH fluctuations of 0.7-1 pH due to the photosynthesis and respiration of seagrass beds, have been reported in different locations 13,15 . Increased ambient pH levels during the day can become locally significant to the point where they have a positive effect on the calcification of co-occurring calcifying algae 12,13 . However, since oceanic conditions are rapidly changing, information is needed about how the presence of seagrasses will affect calcifying algae responses under OA and temperature rise.
Most of the studies regarding the impact of global stressors evaluate the isolated responses of primary producers, using unifactorial models or eventually considering the combined role of OA and temperature rise in the fitness of a specific and isolated biological indicator 16 . Thus far, the expected trend for seagrasses is neutral to positive physiological responses to OA 1 , yet the magnitude of change and affinity for DIC species varies 17,18 . The isolated effects of temperature and CO 2 on the seagrass genus Halodule Endlicher 17,19,20 and their isolated and combined effects on the calcifying green algae genus Halimeda J.V. Lamouroux have been widely addressed [21][22][23][24][25][26][27][28] . The general consensus of OA studies on Halimeda indicates negative to neutral calcification responses and neutral to positive photosynthetic responses to CO 2 -enriched seawater, due to species specificity 22,24,26-33 . To date, two studies have considered the effects of seagrass-calcifying algae interactions under ambient conditions 12,13 , but none have addressed how OA and temperature rise influence these ecophysiological interactions. The species-specific nature of the isolated responses emphasizes the necessity to conduct studies that address OA and temperature rise together in order to better understand the mechanisms behind the presence/absence of interactions between these drivers. Short-term mesocosm experiments that simulate rapid heat waves and acidification, as observed in different regions, are fundamental tools to predict complex ecosystem interactions 34,35 . It is also necessary to introduce realism in these simulations by representing the high-frequency semidiurnal or diurnal variability that dominates coastal or shallow environments 36 . Recent studies reveal that under OA, net photosynthesis of the kelp Ecklonia radiata was almost 50% lower when pH fluctuated than when it was static 37 . This natural variability imposes particularities that can limit or stimulate primary production and must be reproduced in order to properly simulate the predictable future scenarios.
Here we investigate the effects of OA on the photosynthesis and calcification of the seagrass Halodule wrightii and the green alga Halimeda cuneata via a full factorial mesocosm design. We simulate OA and warming by exposing the calcifying alga and the seagrass to the following four combinations of ambient and elevated pCO 2 and temperature: 28 °C & 320 µatm, 28 °C & 822 µatm, 30 °C & 320 µatm and 30 °C & 822 µatm. Most importantly, we examine the degree to which the photosynthetic carbon uptake of H. wrightii influences seawater chemistry under OA through short-term incubations. We aim to determine whether this can act as a metabolic feedback on the photosynthesis and calcification of H. cuneata, considering that these two species of macrophytes coexist in the shallow tropical waters off the Brazilian coast 38 . We hypothesize that H. wrightii is capable of using the excess DIC resulting from OA to increase its photosynthetic activity. In mitigating the effects of OA on seawater

Results
After exposure to treatments for 10 d, short-term incubations (illustrated in Supplementary Fig. S1) of H. cuneata and H. wrightii separately, as well as together, were conducted in order to understand their physiological responses to OA. We also sought to obtain the magnitude of the effect that the primary production of these macrophytes, particularly the seagrass, has on surrounding seawater chemistry under the stress of OA. The ultimate objective was to determine whether biologically altered seawater might be sufficient enough to mitigate the effects of OA on H. cuneata calcification. H. cuneata and H. wrightii fared differently under OA conditions. While the calcifying alga experienced negative physiological consequences (Fig. 1), the seagrass showed a neutral response (Fig. 2, Table 1). We report a significant effect of CO 2 enrichment on H. cuneata calcification (p = 0.017, Table 1), causing it to suffer a 72% decrease under elevated pCO 2 . Simultaneously, we observed a shift in carbonate chemistry when H. cuneata was incubated in the elevated pCO 2 treatment, where HCO 3 − decreased by 56% and total DIC, by 60%, but carbonate and CO 2 remained the same (Fig. 3, Table 2).
We did not detect a significant effect of CO 2 on the gross primary production (GPP) of either H. cuneata (p = 0.370) or H. wrightii (p = 0.321) when incubated separately (Table 1), yet for H. cuneata there was an increasing trend (Fig. 2, Table 2). When H. cuneata and H. wrightii were incubated together, the resulting GPP also did not differ from ambient to elevated pCO 2 ( Table 2), but the overall values were clearly driven by H. wrightii production. The GPP of H. wrightii was about 53 times higher than that of H. cuneata at ambient conditions and 23 times higher under elevated CO 2 ( Table 2).
We observed significant changes in seawater chemistry (Table 3) resulting from the incubation of H. cuneata and H. wrightii separately and together. It's worthy to note, however, that observed shifts that we attribute to an organism's metabolism also reflect the natural chemical equilibrium change that occurs following biological inorganic carbon uptake. The challenge of teasing apart the biological effect and the equilibrium change warrants relative interpretation of DIC uptake. Surprisingly, H. wrightii did not increase DIC uptake under OA, showing no significant differences in total DIC consumption between ambient and elevated CO 2 (p = 0.356, Table 3). However, due to its comparably higher GPP, H. wrightii still took up 40 times more total DIC than H. cuneata

Species CO 2
Seagrass presence   (Table 2). Consequently, when incubated alone under elevated CO 2 , H. wrightii was able to metabolically increase seawater pH by 0.27 ± 0.05 units and aragonite saturation state (Ω Ar ) by 7.57 ± 1.11 units ( Table 2). In contrast, under the same conditions, H. cuneata only increased seawater pH by 0.08 ± 0.04 and Ω Ar by 0.08 ± 0.03 (Fig. 4, Table 2). When incubated with H. cuneata, seagrass presence was a significant factor in determining the evolution of HCO 3 − , CO 3 2− and total DIC as well as ΔΩ Ar and ΔpH in seawater (Table 3). We were not able to quantitatively separate the photosynthetic rates of the alga and the seagrass when they were incubated together. So in order to estimate the effect of the seagrass, we relied on the magnitude of metabolic change for which the seagrass was solely responsible when incubated alone, as mentioned previously.
The seagrass was able to mitigate OA and significantly impact the alga's physiology. We report significant effects of seagrass presence on H. cuneata calcification (p = 0.038), but no interactive effect was found between CO 2 treatment and the latter (p = 0.281, Table 1). When H. wrightii and H. cuneata were incubated together at ambient pCO 2 , there was interestingly no observed change in calcification. However, at elevated pCO 2 , the metabolic interaction between the two mitigated the negative impact of OA and the calcification rate of H. cuneata was reduced by only 34% (as opposed to 72% when alone; Fig. 1, Table 2).
During the experimental period, the ambient CO 2 treatment had a mean pCO 2 of 380 µatm ± 7 SEM and a mean pH of 8.197 ± 0.006 SEM and for the elevated CO 2 treatment, a mean pCO 2 of 822 µatm ± 16 SEM and a mean pH of 7.923 ± 0.007 SEM (Table 4). pH fluctuated throughout the experiment due to natural variation from the adjacent reef, as shown in Supplementary Fig. S2. Mean A T was 2278 ± 2 (ambient) and 2280 ± 7 (elevated; Table 4). There was no effect of CO 2 treatments on mean seawater TA (p = 0.723). The ranges of these parameters as well as the remaining physicochemical characterisation of seawater are found in Table 4. Temperature showed no significant effects or interactions on all descriptors (Table 1). Supplementary Fig. S3 shows the average PAR values observed throughout the day during the experimental period.

Discussion
The presence of the seagrass H. wrightii mitigated the negative effect of OA on the calcification of the alga H. cuneata. This is the first study to confirm that under elevated CO 2 concentrations, seagrass is still capable of maintaining comparatively high photosynthetic rates, and in turn, creating seawater conditions that are conducive to the calcification of sympatric, and otherwise ill-fated, calcifying algae. On their own, H. cuneata and H. wrightii responded differently to OA. The alga suffered decreased calcification and both the alga and the seagrass showed   no significant photosynthetic response. It has widely been shown that calcifying organisms respond negatively to OA, whereas fleshy plants respond neutrally or positively 1,39 . The difference in the magnitude of the metabolic effect that each organism had on the surrounding seawater was substantial, where the dominant effect of H. wrightii played to the calcifying alga's advantage under OA. It is worthy to note that the range of temperature initially tested (28-30 °C) to simulate ocean warming need not be considered a frontrunner threat to H. cuneata and H. wrightii photosynthesis and calcification due to the absence of an observable effect of temperature on these processes.
One of the more interesting findings of our study is the apparent asymmetry between photosynthetic and calcification responses to OA in H. cuneata. The calcified macroalgae showed signs of physiological stress, seeing as we observed a substantial decrease (72%) in its calcification. On the other hand, the photosynthetic response was not significant due to variation within treatment, however the fact that there was a 54% increase is worthy of consideration. These two processes occur side by side at the cellular level. Calcification occurs in intercellular spaces (inter-utricle spaces, or IUSs), which are separated from bulk seawater by a layer of utricles where photosynthesis is concentrated 40 . DIC uptake during photosynthesis is found to increase the pH of IUSs and calcium carbonate precipitation is favoured. Conversely, calcification produces H + and CO 2 , which balance the change in pH and CO 2 concentration produced by photosynthesis 40 . Due to the close proximity of these processes, under ambient conditions, calcification is reported to be closely coupled to photosynthesis in the genus Halimeda 41,42 . Thus, when photosynthesis increases, calcification is expected to increase, and vice versa. However, when we elevated CO 2 in this study, calcification was compromised despite an increasing trend in GPP (Table 2), which suggests that there may be a certain degree of independence between these processes.
The understanding of this uncoupling lies in the carbonate chemistry dynamics of the location where photosynthesis and calcification intersect, the IUSs. Ultrastructure data suggests that the structure and size of utricles and IUSs in Halimeda may help to explain the carbonate chemistry of the IUSs and thus, OA responses 24,41 . Peach et al. 2017 found an inverse relationship between diffusive pathway type and mineral content, where species with longer utricles and thinner pathways contained more aragonite than those with shorter utricles and wider pathways. Morphological parameters were not one of our response variables, but based on our results and previously established calcifying mechanisms for Halimeda 40,43 and aquatic plants 44 , we suggest that the diffusive pathway of H. cuneata permits corrosive bulk seawater to replenish the IUSs at a much faster rate than it can be biologically-regulated, thus partially inhibiting calcification 45 . We also suggest that dissolution may be   Table 4. Seawater characterisation of the two CO 2 treatments. Shown are the means ± SEM (and ranges for pH and pCO 2 ) for each parameter.
contributing to the low pH and DIC-rich environments in the IUSs. The degree to which the alga may be experiencing dissolution is also not evident, due to the difficulty of disentangling the effects of dissolution and decreased calcification in OA studies 42,46 . The alga may be using DIC directly from dissolution as substrate for photosynthesis, thus explaining why DIC from bulk seawater was in lesser demand, shown by the 60% decrease in DIC consumption. However, the GPP of the alga was not capable of ameliorating the imbalance in carbonate chemistry of IUSs enough to stimulate calcification, thus our data supports the hypothesis that photosynthesis and calcification become uncoupled under OA 1 . Nonetheless, we cannot be certain of the source of DIC for this increasing trend in photosynthesis, nor the reason that the alga was unable to further increase GPP. Further research on this species using microsensors would be essential to ascertain these unknown thresholds that explain the apparent disparity between photosynthetic and calcification responses to OA 46 . We initially expected an increase in the GPP of H. wrightii, since additional CO 2 substrate is expected to stimulate primary production in fleshy marine plants. However, our results show that there was no significant change in photosynthesis at elevated CO 2 . Recent meta-analyses report neutral to positive photosynthetic responses of seagrasses to elevated CO 2 1,47 , which maintains our results within the range of expected responses. In addition, a study that exposed tropical H. wrightii to reduced pH observed an increase of only 20% in its photosynthetic rate, followed by a prominent plateau that was attributed to a preference for HCO 3 − use 17 . In the same study, the absence of a change in photosynthesis in H. wrightii after the exposure to acetazolamide (AZ), an inhibitor of carbonate anhydrase (CA), a common enzyme that aids in conversion of HCO 3 − to CO 2 , indicated that this species has an alternate and more efficient mechanism for HCO 3 − use when compared to other seagrasses 17 . Our observations likely indicate that H. wrightii's neutral response to OA is due to the efficiency of its mechanism of HCO 3 − use. Due to its much higher photosynthetic rate (464 ± 60 (SEM) µmol O 2 gDW −1 h −1 ), when compared to H. cuneata (19.5 ± 2.7 (SEM) µmol O 2 gDW −1 h −1 ) at elevated pCO 2 , H. wrightii removes 40 times more DIC from seawater, thus increasing the pH 48 , CO 3 2− availability and aragonite/calcite saturation states 49 . Although the seagrass did not increase its GPP, its capacity to biologically alter its surrounding seawater chemistry was enough to influence the metabolism of coexisting H. cuneata. We did not quantify the density of the studied Halodule bed, which is a factor that is shown to affect the magnitude of Halimeda-seagrass interactions 12,50 . Our results show that Halodule populations are likely to withstand intermediate OA scenarios, yet local irradiance, temperature and nutrient conditions may very well play a determinant role in the magnitude of the metabolic interactions between seagrasses and sympatric calcifying macroalgae. Interspecific variations in seagrass photosynthesis due to diverse DIC assimilation mechanisms will also put some species at an advantage over others 17,51 . This was observed at volcanic CO 2 vent sites, where seagrass community composition shifted according to seawater pH 52 . The extent to which populations are acclimated to elevated conditions may determine their long-term resilience.
Acidified seawater is ultimately unfavourable for H. cuneata calcification, however we demonstrate that high-performing primary producers such as H. wrightii are capable of providing significant refuge for these calcifying algae via biologically altering seawater chemistry. Previously, a 1.6-fold increase in the calcification rate of Halimeda renchii was observed in seagrass beds at ambient CO 2 levels 13 . In our study, it was unexpected that seagrass presence did not also increase H. cuneata calcification at ambient CO 2 . Based on the high GPP of H. wrightii observed at ambient CO 2 , one would expect the consequent IUS carbonate chemistry to be exceptionally favourable and to stimulate calcification. Regardless, the issue is that future oceans will possess a much higher pCO 2 than that of today's oceans. Our results show that under OA, the presence of seagrass will likely foster calcification rates during the day that are comparable to those at current pCO 2 . Recent findings anticipate, however, that other factors of different functional scales will cause variation in this buffering capacity. There are often other marine macrophytes coexisting with Halimeda and seagrass, namely macroalgae. Benthic community composition is known to alter seawater chemistry at different magnitudes 53,54 , greatly due to species-specific irradiance optima and CCM mechanisms, therefore influencing the community's OA buffering capacity. Modeled projections incorporating effects of OA and net community metabolism (NCM) on carbonate chemistry in seagrass meadows predict long-term offsets of CO 2 , but also NCM-driven extremes in carbonate chemistry under OA 55 . In particular, future pH levels at night are expected to be extremely low due to the intensified effect of OA on respiration 55 . This has implications for the net calcification of Halimeda that we weren't capable of addressing and would need to be analysed in future studies. Additionally, Cyronak et al. (2018) reveal that the spatial and short temporal variation of carbonate chemistry in seagrass beds can be even greater than diel variability, thus potentially impacting the buffering capability of seagrasses across even smaller scales 56 . The fate of calcifying algae under OA may very well lie in the composition of the accompanying photoautotroph community as well as their associated NCM dynamics 54 .
Solid predictions of whether and which calcifying algae will adapt to OA & temperature rise are generally still insufficient 57 , partially due to the lack of incorporation of species interactions effects and natural seawater variability. Most of the available information produced until now has been based on stable values of pH and temperature 37 and few global studies exist that address marine plant interactions alongside OA and temperature rise 58 . Despite the academic value of these efforts, their utilization in depicting future scenarios should be considered with caution, since natural variability of physical/chemical conditions is a selective pressure and a major driver of marine ecosystem functioning. Likewise, although there are known limitations to not manipulating CO 2 directly into each experimental tank 59 , our design was chosen as the most feasible, which gave priority to the incorporation of diel pH and CO 2 variability. Each tank was an isolated experimental unit with a certain degree of intrinsic variability and our results do not suggest that our design has biased the experimental outcome.
The degree of physiological tolerance or increased performance to changes in CO 2 and temperature in the marine environment can be due to trans-generational plasticity, phenotypic buffering, or plasticity within generations (or 'classical' plasticity) from which 'true' evolutionary adaptation may arise 6 . Data supports that genetic variation in traits important for OA and temperature rise is prevalent in near-shore plants 6 . Based on our results, we suggest that the large natural variability of temperature and CO 2 in shallow coastal environments has selected Scientific RepoRts | (2019) 9:1932 | https://doi.org/10.1038/s41598-018-35670-3 for phenotypic plasticity and co-evolutionary tools involving the metabolic interaction between H. cuneata and H. wrightii, thus potentially providing resilience and adaptability to OA. Plasticity in response to OA and temperature rise will help maintain population resilience under changing environments 60 . If Halimeda species have adequate genetic variability to generate phenotypes with different CO 2 tolerances and optima, then it is likely that inter or intraspecific variability in fitness will be observed, where OA winners are likely to be those coexisting with seagrasses. The responses we observed are a contribution to the understanding of possible shifts in composition of relevant communities 61 , but they also highlight the relevance of coastal plant metabolic interactions as a dynamic biological factor that should be considered in the management of natural habitats, namely marine protected areas, in view of future climate scenarios.

Methods
Study area and experimental design. The experiment was performed in a large-scale, flow-through mesocosm designed by Projeto Coral Vivo, located at its research station on Araçaípe beach, Bahia, Brazil. The mesocosm was designed to test the effects of ocean acidification and warming (among other factors) on reef organisms 62 , while closely mimicking the adjacent reef conditions. Aracaípe beach's (16° 29′ 28.6′′ S 39° 3′ 58.4′′ W) fringing reef develops 100 m off the coast of the Marine Mesocosm, where the seagrass H. wrightii and the upright calcifying green alga H. cuneata coexist. The open flow and proximity of the mesocosm to the fringing reef make this system highly realistic since it is able to maintain experimental conditions (seawater composition, temperature, diel pH and CO 2 variability, turbidity, salinity, plankton density, photoperiod, rainfall, irradiance…etc) that are very similar to those in the adjacent reef.
With the interest of simulating moderate predictions of ocean acidification and warming, levels of pCO 2 and temperature were chosen based on the IPCC RCP 6.0 scenario 63 and modeling of future atmospheric emissions 64,65 . We initially established a full factorial design of ambient temperature and pCO 2 and elevated temperature and pCO 2 , totaling four treatments. The ambient levels were the unaltered present temperature (28 °C) and pCO 2 (380 µatm) of seawater from the adjacent reef, and the elevated CO 2 and temperature treatments were achieved by manipulating seawater to target +2 °C (30 °C) and +0.25 pH, or +442 µatm (822 µatm). The pCO 2 was calculated for each treatment using the TA and mean pH values via CO2SYS 66 .
Seawater from 500 m offshore was continuously pumped into four 5,000-L underground sumps where the four CO 2 and temperature treatments were applied. pCO 2 was manipulated in two sumps using a custom reactor system that introduced fine bubbles of CO 2 into constantly mixed seawater. Similarly, a 1.9 m 15,000 W heater was placed in each of the two underground sumps where seawater temperature was to be elevated. Seawater pCO 2 and temperature were not fixed. We used a custom-made Reef Angel Open-Source Controller, which elevated and regulated pCO 2 and temperature levels with respect to ambient fluctuations. Treatments were applied to header sumps and not directly to experimental tanks based on feasibility and limitation of resources. Mixed treatment water was fed to four 310-L reservoir tanks, from which flow was regulated to 16, fully randomized 130-L raceway experimental tanks (n = 4 per treatment). Experimental tanks were continuously supplied with seawater at a flow rate of ~10 L•min −1 , achieving a renewal rate of 5 x per hour. The experimental tank area was covered in shade cloth to uniformly reduce the intensity of natural sunlight by 70%, simulating the amount of incident solar radiation measured at approximately 2 m where organisms were collected on the reef. Duarte et al. (2015) reported a complete description of the mesocosm design and functioning.
Sampling. Approximately 160 specimens of H. cuneata and 1,500 shoots of H. wrightii were collected at a depth of 2 m using SCUBA by carefully removing the entire holdfast and rhizome, respectively, and were brought to the holding aquariums of the Marine Mesocosm for sorting and removal of epibionts. Sediment from the first 10 cm of the sampling area was also collected and used as substrate for subsequent planting in the mesocosm. Ten H. cuneata thalli were placed upright in a plastic tray (40 × 17 × 4 cm), with the holdfasts anchored in 3 cm of sand. One tray was placed in each of the 16 experimental tanks. Approximately 50 seagrass shoots were replanted in each of 2 plastic trays with 3 cm of sediment in each experimental tank. Each of the 16 experimental tanks thus possessed 3 trays, 1 with H. cuneata and 2 with H. wrightii. Organisms were acclimated at ambient temperature and pCO 2 for 15 days. Treatments commenced upon completion of the acclimation period, reaching target levels within 24 hours and were applied for a total of 10 days.
Abiotic parameters. Salinity (Refractometer: Instrutherm RTS-101ATC), dissolved oxygen & temperature (Portable dissolved oxygen meter: Instrutherm MO-900), incident irradiance (Quantometer: apogee MQ-200) and pH (pHmeter: Gehaka ISO 9001) were measured daily in each experimental tank. Handheld pH meter and pH sensors were calibrated to NBS buffers daily and sensor drift was checked weekly using a bench top Gehaka pH meter. The remaining abiotic parameter meters were calibrated as per recommended in their factory manuals, using appropriate calibration solutions. The daily average photosynthetically active radiation (PAR) was monitored with light loggers (HOBO), which were positioned underwater at the level of the organisms in the experimental tanks. Nutrient concentrations were monitored every 3 days in each tank 67 . For the monitoring of seawater carbonate chemistry, water samples were retrieved from each experimental raceway tank (n = 3) and were immediately refrigerated. Total alkalinity measurements, were performed using a custom USB4000 spectrophotometer (Ocean Optics, Dunedin, USA) and compared to certified reference material (Scripps Institute of Oceanography, USA).
Primary production and calcification. In order to isolate and assess the potential metabolic interactions between H. cuneata and H. wrightii, short-term (2.5 h) incubations were administered at the beginning and end of the experiment. Incubations were conducted on each species separately, as well as with the two species together Scientific RepoRts | (2019) 9:1932 | https://doi.org/10.1038/s41598-018-35670-3 (n = 3 per species/combination). Oxygen evolution and change in total alkalinity were measured; the former was used to calculate gross primary production (GPP) and the latter, for calcification rates and change in dissolved inorganic carbon (ΔDIC). GPP and ΔDIC were calculated for each species, whereas calcification was only calculated for the alga. An irradiance of 750 µmol quanta m −2 s −1 was used during the incubations since this was the average midday irradiance during the time period when the incubations were conducted.
Incubations did not take place in the experimental raceway tanks. An incubation setup was constructed outside of the experimental tanks using 28-L boxes, impermeable plastic bags as chambers and an illuminator. The illuminator structure was equipped with four metallic vapor lamps (220 V, REV426A4, Serwal) and positioned over four independent dark 28L boxes (22.5 × 35 × 50 cm, Marfinite). Each box was connected to an individual raceway experimental tank by a 12 mm (diam.) hose so that seawater was constantly renewed in the box. Thus, each box corresponded to a treatment (Supplementary Material Fig. S1). In each box, four transparent 29 × 39 cm incubation bags impermeable to dissolved oxygen served as incubation chambers and were filled with approximately 500 ml of seawater. Three of the bags received the following biological material: (1) only 1-2 H. cuneata thalli, (2) 1-2 H. cuneata thalli and 10 shoots of H. wrightii or (3) only 10 shoots of H. wrightii (illustrated in Supplementary Material Fig. S1). The fourth bag contained only seawater in order to monitor any background changes in oxygen concentration and total alkalinity (A T ) due to microorganisms. Water samples were taken directly from treatment boxes for initial measurements of dissolved oxygen concentration (DO) and A T immediately prior to commencement of incubations. All visible air bubbles were removed from the bags before their sealing. At the end of the incubation period, water samples were taken from each bag for the final measurements of DO and A T . Water volume was measured and H. cuneata and H. wrightii were removed, dried at 60 °C and weighed. Basal segments were removed from each sample, decalcified with nitric acid (0.6M HNO 3 ) and weighed to determine the dry decalcified weight.
Oxygen evolution. Five initial water samples (12 ml) were collected from each treatment box (n = 3) at the beginning of the incubation. At the end of the incubations, five water samples were taken from each bag (n = 3) with a 60 ml syringe fitted with a small tube. Samples were immediately treated with manganese chloride and alkaline-iodide reagents upon removal and refrigerated for 72 hours. They were then treated with a sulfuric acid reagent and analysed spectrophotometrically according to the Winkler method adapted by Labasque 68 in order to calculate the dissolved oxygen concentration, or O 2 production. GPP values were calculated by normalizing O 2 production to incubation time, volume of water and the decalcified dry weight of the incubated tissue (µmol O 2 gDW −1 h −1 ), after removing background O 2 fluctuations due to microbial activity. Calcification. One initial water sample (180 ml) was collected from each treatment box pre-incubation and one final water sample was taken from each bag post-incubation (n = 3). Water samples were immediately refrigerated until analysis. Alkalinity anomaly measurements were performed using the aforementioned Ocean Optics equipment. The CO2SYS program 66 was used to calculate all DIC species and components of seawater carbonate chemistry. Changes in each DIC species (ΔHCO 3 − , ΔCO 2 and ΔCO 3 2− ) and total DIC (ΔDIC) were calculated by subtracting the pre-incubation value from the post-incubation value.
Calcification rates were calculated for H. cuneata using the following equation 69 : (1) T where g = µmol CaCO 3 g −1 h −1 , ΔA T = change in total alkalinity, V = volume of incubated seawater, DW = dry, decalcified weight of H. cuneata and Δt = incubation time (h). statistical analysis. A three-way analysis of variance (ANOVA), was performed on the H. cuneata calcification and GPP data (log(x + 1) transformed), with the factors CO 2 (two levels), temperature (two levels), and seagrass presence (two levels). A two-way ANOVA was used to test H. wrightii GPP data (log(x + 1) transformed), with the factors CO 2 and temperature (two levels each). Significance level was set at p = 0.05. Due to the absence of any temperature effect and a strong trend in the data with respect to CO 2 , we proceeded to pool the samples from the same CO 2 treatment for a more robust analysis and in order to preserve important ecological implications. All data passed assumptions of normality of residuals and homogeneity of variances. Subsequently, two-way ANOVAs were applied to H. cuneata calcification, GPP data (log(x + 1) transformed), ΔHCO 3 − , ΔCO 2 , ΔCO 3 2− , ΔΩAr, ΔpH and Δtotal DIC data (log(x + 1) transformed), with the factors CO 2 (two levels) and seagrass presence (two levels). One-way ANOVAs were applied to H. wrightii GPP data (log(x + 1) transformed), ΔHCO 3 − , ΔCO 2 , ΔCO 3 2− and Δtotal DIC data (log(x + 1) transformed), with CO 2 (two levels) as a factor. Fisher's LSD post hoc tests were used for pairwise comparisons of significant effects. All statistical analyses were performed using IBM SPSS Statistics 24.

Data Availability
The datasets generated during and/or analysed during the current study are available from the corresponding author upon reasonable request.