Linking gene expression to productivity to unravel long- and short-term responses of seagrasses exposed to CO2 in volcanic vents

Ocean acidification is a major threat for marine life but seagrasses are expected to benefit from high CO2. In situ (long-term) and transplanted (short-term) plant incubations of the seagrass Cymodocea nodosa were performed near and away the influence of volcanic CO2 vents at Vulcano Island to test the hypothesis of beneficial effects of CO2 on plant productivity. We relate, for the first time, the expression of photosynthetic, antioxidant and metal detoxification-related genes to net plant productivity (NPP). Results revealed a consistent pattern between gene expression and productivity indicating water origin as the main source of variability. However, the hypothesised beneficial effect of high CO2 around vents was not supported. We observed a consistent long- and short-term pattern of gene down-regulation and 2.5-fold NPP decrease in plants incubated in water from the vents and a generalized up-regulation and NPP increase in plants from the vent site incubated with water from the Reference site. Contrastingly, NPP of specimens experimentally exposed to a CO2 range significantly correlated with CO2 availability. The down-regulation of metal-related genes in C. nodosa leaves exposed to water from the venting site suggests that other factors than heavy metals, may be at play at Vulcano confounding the CO2 effects.


Results
Seawater parameters. A description of the seawater parameters and irradiance during the incubations conducted both in Vulcano (Italy) and Ria Formosa (Portugal) is given in Table 1. In Vulcano, mean pCO 2 concentrations (and associated parameters of the DIC system) were significantly higher in the CO 2 site than in Reference site (p < 0.05, df = 6). In accordance with descriptions from previous studies 29,30,34 conducted in this site, seawater chemical variability near the venting site (CO 2 site) was higher than in the Reference site, particularly regarding the inorganic carbon system. Indeed, occasional extreme high values of pCO 2 and DIC were recorded in the CO 2 site (Table 1).
Scientific RepoRts | 7:42278 | DOI: 10.1038/srep42278 Vulcano Incubations. Gene expression. Expression stability of reference genes (RGs): Four RGs (Eukaryotic initiation factor 4A (eIF4A), Glyceraldehyde 3-phosphate dehydrogenase (GAPDH), 18S ribosomal RNA (18S) and Ubiquitin (UBI)), belonging to different functional classes, were tested for stability in the different experimental conditions ( Table 2). The three algorithms applied (BestKeeper, geNorm and NormFinder) agreed in suggesting the eIF4A as the best RG (see Supplementary Tables S1, S2 and S3 in Appendix 1, Supplementary Information). However, we decided to include GAPDH as RG, indicated by geNorm, and also the 18S, the second best RG according to NormFinder and geNorm, for target gene expression normalization (Supplementary Table S4). The use of three RGs allowed for a more reliable normalization of gene expression data.
Genes of interest (GOIs) expression level: Relative expression levels of 20 GOIs involved in photosynthesis, carbon dioxide fixation and metabolic carbon assimilation pathways, antioxidant and metal responses were analysed in C. nodosa plants from both Reference and CO 2 sites after in situ and transplanted incubations ( Table 2).
Major changes in gene expression patterns were observed in plants long-term growing near the vents (in situ incubations in CO 2 site) with respect to plants from the Reference site (in situ incubations in Reference site) (Fig. 1a), as confirmed by the multivariate analysis of similarity (ANOSIM), which revealed a certain degree of separation between the two sites (Global R: 0.370, p = 0.098, Supplementary Table S5). Overall, expression levels of genes involved in light-dependent reaction of photosynthesis, carbon dioxide fixation and carbon assimilation pathways, were reduced in plants long-term growing in the CO 2 site compared to plants from Reference site (Fig. 1a). According to univariate REST analysis, a significant down-regulation was observed for the subunit psbA of the Photosystems II, the Photosystem I Light Harvesting Complex gene 1 (LHCA1), the Beta carbonic anhydrase (BCA), and the Sucrose synthase (SUS) (P(H1): 0.000), as well as the large subunit of RuBisCO (rbcL) (P(H1): 0.037). The transcripts for the Ferredoxin (FD) and Phosphoenolpyruvate carboxylase (PEPC) were close to the significance level (P(H1): 0.058 for both) (Supplementary Table S6). The same pattern was found for genes involved in the antioxidant response and metal detoxification (Fig. 1a). The expression of genes involved in reactive oxygen species (ROS) detoxification enzymes (antioxidant system), such as Superoxide dismutase (SOD) and Ascorbate peroxidases (APX7 and APX6), was significantly reduced, as were genes involved in glutathione metabolism (GSH-S and GR) and luminal binding protein (LBP) (P(H1): 0.000). Down-regulation of metal tolerance and metal detoxification proteins (MTP and MT) was also significant (P(H1): 0.000). The general down-regulation pattern observed points to an overall reduction of metabolic processes related with photosynthesis, oxidative (ROS) and metal detoxification response in plants growing near the CO 2 vents with respect to plants from the Reference site.
When specimens of C. nodosa from the Reference site were incubated in water from the CO 2 site (transplant incubation), expression of almost all the GOIs analysed was reduced (Fig. 1b) also indicating an overall reduction of metabolic processes related with photosynthesis and antioxidant response when compared to in situ conditions (i.e. plants from the Reference site incubated in water from the Reference site). Significant down-regulation was also detected for the enzymes Phosphoenolpyruvate carboxylase (PEPC) and Glutathione reductase (GR) (P(H1): 0.000) (Supplementary Table S6).
On the contrary, when plants growing in the CO 2 site were incubated in water from the Reference site, a clear gene up-regulation pattern was observed on the metabolic paths linked to photosynthesis as well as oxidative (ROS) and metal detoxification to acclimate to the new conditions (Fig. 1c). Although multivariate differences in gene expression were not significant (Global R: 0.370, p = 0.200, Supplementary Table S5), univariate REST analysis revealed a significant up-regulation of the photosynthesis-related genes psaJ, and rbcL, as well as the antioxidant-related genes CAT, APX6 and GSH-S (P(H1): 0.000) (Fig. 1c). Several GOIs analysed (psbA, L, LHCA1, SUS, BCA, APX7,GR, MTP) were also close to the significance level (P(H1) < 0.1) (Supplementary Table S6).
Net plant productivity. In Vulcano, the net plant productivity (NPP) of C. nodosa was significantly affected by the origin of the water where plants were incubated (p < 0.01). NPP was significantly lower when plants were  (Fig. 2). Plants collected from both sites did not differ significantly in their exposure to different CO 2 concentrations in water (i.e. no significant interaction "plant origin" x "water" was detected, F 1,25 = 2.698, p = 0.113, see ANOVA results in Supplementary Table S7).
Gene expression and productivity. A PCA and cluster analysis and a correlation analysis were performed to link gene expression and productivity results. The PCA and Cluster analysis, conducted considering the combined contribution of all GOIs analysed and NPP, clearly highlighted two groups (PCA component 1 explaining 66.5% of the total variance) defined by the origin of the water where plants were incubated (i.e. water from Reference vs. CO 2 site) (Fig. 3). These two groups matched the patterns described in Figs 1 and 2. A general up-regulation of photosynthetic, antioxidant and metal detoxification-related genes and higher NPP was found in plants long-term growing in the Reference site (in situ incubations) and plants from CO 2 site incubated in water from the Reference site (transplant incubations). Reversely, plants long-term growing in the CO 2 site (in situ incubations) and plants from the Reference site incubated in water from the CO 2 site (transplant incubations) showed a generalized gene down-regulation and lower NPP. The biplot in the PCA identify GR, MT, PEPC, SOD, GSH-S, LHCA1 and psbA as the genes most contributing to this separation (for detailed information on the contribution of all GOIs and NPP see Supplementary Table S8).

Ria Formosa Incubations.
In Ria Formosa, where there are no volcanic vents, a set of in situ incubations of C. nodosa plants were performed under the same physico-chemical conditions with CO 2 levels experimentally controlled ( Table 1). NPP of C. nodosa plants in the Ria Formosa incubations showed a significant positive correlation with DIC (adjusted R 2 = 0.42, p < 0.01) (Fig. 4).

Discussion
Our results showed that both the net plant productivity (NPP) and the gene expression of C. nodosa were significantly lower in plants incubated in CO 2 -rich water from Vulcano CO 2 vents when compared to Reference water. Plants living in the venting site showed a general reduction in the expression of genes involved in various metabolic processes, particularly those related with the light-dependent reactions of photosynthesis, carbon fixation and metabolic carbon assimilation when compared to plants from the Reference site. The same pattern was observed for genes involved in the oxidative stress antioxidant response and metal detoxification. This generalized gene down-regulation was paralleled by a reduction in plant metabolism, which resulted in 2.5 fold decrease of NPP in plants naturally growing under acidified water near the vents. Broad-scale decreases in the expression of genes related to key cellular processes, including metabolism and stress response, affecting physiological processes have been already described in different marine taxa in response to ocean acidification (ref. 35 and references therein). During long-term CO 2 enrichment, down-regulation of RuBisCo has been described in C 3 plants as a photosynthetic acclimation response (ref. 25 and references therein). A reduction in global metabolism has been also described in many marine organisms under high CO 2 conditions 8 . A general down-regulation of photosynthesis-related genes has been described in seagrasses of the genus Zostera in response to temperature increases and global warming 17 , but current knowledge on seagrasses gene expression patterns under high CO 2 conditions is still very limited. The only other available study described variable patterns on expression of stress-related genes in the seagrass Posidonia oceanica growing near the volcanic CO 2 vents of Panarea and Ischia Islands (Italy) 22 . In agreement with our results, these authors reported down-regulation of several metal-detoxification and antioxidant genes and suggested there was no need for activation of these metabolic defence mechanisms due to the low heavy metal-stress and oxidative damage. Seagrass species show many convergent adaptive features to allow life in the marine environment (see refs 36 and 37), however different lineages can have different response to specific biotic and abiotic factors. We have analysed and identified reference genes (RGs) that are stable in C. nodosa exposed to volcanic CO 2 vents and the expression level of the genes of interest (GOIs). The pioneer information reported here can represent a baseline for understanding adaptive features and stress responses in this species and in the Cymodoceaceae family in general.
The short-term acclimation responses of C. nodosa to high CO 2 , evaluated through the transplant incubations, agreed with the molecular and productivity long-term responses measured in situ. Reference plants incubated in water from the CO 2 site experienced a generalized, drastic down-regulation in the gene expression of several photosynthetic, antioxidant and metal-related genes that globally reduced plant metabolism and resulted in a significant NPP decrease. On the contrary, plants from CO 2 site incubated in Reference water significantly up-regulated the expression of those genes and moderately increased NPP.
Globally, we identified a coordinated gene expression pattern among photosynthetic antioxidant and metal-detoxification genes in all incubations conducted in Vulcano. We also detected a fast response of gene expression over physiology and global metabolism in C. nodosa leading to significant differences in plant productivity when incubated in water with different CO 2 concentrations. Significant positive correlations were detected between gene expression and productivity for specific genes encoding proteins related with the efficiency of the photosynthetic process, such as the Photosystem I reaction center (psaJ) and the Photosystem II protein D1 (psbA), the Photosystem I light harvesting complex (LHCA1), and the carbon fixation (Phosphoenolpyruvate  carboxylase, PEPC). Significant correlations were also detected between oxidative stress-related genes (Superoxide dismutase, SOD, Glutathione reductase, GR, and Glutathione synthase, GSH-S) and productivity. These results support the use of molecular tools as early indicators of potential stress or physiological readjustments in response to environmental change 38 since metabolic pathways and physiological machinery are primarily gene-coordinated 10 . They may thus provide earliest evidence of organism responses before than morphological and physiological indicators 39 . Work is still needed to untangle the scaling-up control from genes, over protein expression and biochemistry up to physiology in seagrasses, their energetic implications and the metabolic consequences.
The contrasting response pattern recorded in Vulcano between plants long-term growing at each site (in situ incubations) could be attributed to local adaptation and differential selection of specific genotypes under acidified conditions 40 . However, preliminary results of a population genetic analysis carried out in Vulcano using microsatellite markers, at the same time and sampling sites than the present study, showed no genetic differentiation and high gene flow between plants from CO 2 and Reference sites (Silva et al. in prep.). Consequently, results point to a low contribution of local adaptation to the observed differences between plants from the two sites. Organisms may also be able to cope with conditions found within the CO 2 vents through plastic responses at different levels, from molecular (i.e. gene expression) to physiological, morphological and behavioural, without genotypic changes occurring through selection 41 . The large photosynthetic plasticity attributed to C. nodosa, in terms of photosynthetic response capacity 42,43 , could thus explain the differences found between specimens in Vulcano sites and the rapid acclimation mechanisms triggered during the transplant incubations.
The observed decrease of C. nodosa productivity under the influence of the high CO 2 vents in Vulcano island appears to contradict the common assumption that seagrasses will increase their productivity in a foreseen high CO 2 world. Contrastingly, the pattern obtained in Ria Formosa incubations, where CO 2 availability was the only factor tested, significantly correlated with DIC availability. This latter pattern supports the hypothesis that seagrasses are C-limited at current CO 2 concentrations 4,32,44 and that they may benefit from higher DIC availability, enhancing photosynthetic rates and productivity if no other limiting factors, such as irradiance, nutrients or temperature, are present 7 . The photosynthetic plasticity attributed to C. nodosa 42,43 could once more explain, via acclimation to local conditions, the differences found between specimens from distant locations long-term growing under different environmental pressure (i.e. high CO 2 exposure in Vulcano vs. present day CO 2 exposure (ca. 380 μ mol pCO 2 ) in Ria Formosa) when exposed to high CO 2 . However, the sharp NPP decrease observed in C. nodosa growing at present day CO 2 levels in Vulcano (Reference plants) when incubated in CO 2 -rich seawater (i.e. transplants) is difficult to justify, particularly when considering the photosynthetic carbon-limited physiology of this species 44,45 and the significant correlation of NPP to CO 2 showed by Ria Formosa plants short-term exposed to a CO 2 range.
The PCA analysis revealed a consistent pattern between gene expression and productivity indicating water origin as the main source of variability among all the incubations conducted in Vulcano. Results point to the possibility that other environmental factors related to volcanic vent emissions may correlate with the high CO 2 in Vulcano 46 . Since plants from both CO 2 and Reference sites in Vulcano were growing and incubated under similar environmental conditions (i.e. irradiance, temperature, salinity (Table 1) and nutrients 47 ) only differing in their exposure to the venting seeping fluid, its specific physico-chemical composition may be the cause of the particular response of C. nodosa. Submarine CO 2 vents share the common trait of CO 2 being the main gas emitted. However, the particular physico-chemical composition of seeping fluids is site specific and varies widely among vents 29 . The seeping fluid in Vulcano has been characterized as a CO 2 source (97-99%) with sulphide (H 2 S) concentrations of about 270 μ M 48 and minor concentrations of trace elements and toxic metals (e.g. Cd, Co, Cr, Cu, Mn) 46 . Negative and toxic effects of H 2 S and heavy metals on the photosynthetic apparatus, metabolism and survival of marine organisms, including seagrasses, have been previously described 49,50 . The down-regulation of metal detoxification genes (MTP and MT) recorded in C. nodosa leaves when incubated in water from CO 2 site suggests no significant damage associated with heavy metals. However, attenuation, and even inhibition, of cytochromes functioning has been reported at concentrations of H 2 S as low as 100 μ M compromising the photosynthetic output of photosystems 51,52 . This is a likely explanation for the significant down-regulation detected in the photosynthesis-related genes, particularly in the photosystems I and II, and the lower productivity of C. nodosa plants incubated with water from the venting site in Vulcano. Unfortunately, no data on seawater sulphide or metal concentration during the experimental period are available. At present, a major knowledge gap exists regarding the combined effects of high CO 2 and other drivers associated with Global Climate Change on marine benthic communities, including those dominated by seagrasses. The particular chemical composition of seeping fluids at each venting site may recreate different foreseen multi-drivers scenarios associated with Climate Change (i.e. hypoxia, high temperature, sulphide and metal presence, anthropogenic pressure) 1 .
In conclusion, we related, for the first time, gene expression with net plant productivity in the seagrass species, Cymodocea nodosa, exposed to high CO 2 in volcanic CO 2 vents. We found that the gene expression and productivity responses of C. nodosa to high CO 2 at Vulcano vents were coupled indicating water origin as the main source of variability among all the incubations. However, the hypothesis that high CO 2 environment near the vents is beneficial to plant productivity was not supported. The consistent and unexpected long-and short-term pattern of gene down-regulation and NPP decrease observed in C. nodosa when incubated with water from the vent site suggests that environmental factors, other than increased inorganic carbon availability, may be at a play affecting photosynthetic metabolism and decreasing productivity. The down-regulation of metal detoxification genes suggests no significant damage associated with heavy metals. Further research is needed to clarify if sulphide or other components of seeping fluids may explain our observations. Natural submarine CO 2 vents provide unique experimental conditions to evaluate the acclimation and adaptive potential of species and communities to scenarios where ocean acidification occurs in combination with other environmental factors.

Material and Methods
Site Description. This study was conducted in May 2013 in Vulcano Island (Sicily, Italy) where submarine gas emissions date from late 18 th century 53 . The experimental site was located in Levante Bay, a shallow bay under the influence of a submarine volcanic vent 46 . A pH gradient is observed along the coast, with the lowest values near the venting emission points (pH 5.65) increasing to normal values (pH 8.1) at about 400 m away from the vents 30 . Fluctuations in DIC and pH along this gradient have been described depending on wind direction and consequent water mass movement 30 .
We selected two sites, characterized by different pH regimes, where the seagrass Cymodocea nodosa grows 30,34,46 . The low pH/high CO 2 site (CO 2 ) is located approximately 200 m from the vent at 2 m depth and 50 m from the shore. The Reference site (REF), away from the vent influence, is located approximately 450 m from the vent at 2.5 m deep and 40 m from the shore.
In order to test the productivity results obtained in Vulcano Island, we conducted a set of incubations at an independent control site. Specimens of C. nodosa growing in Ria Formosa (Portugal), a lagoon system with no venting influence, were incubated following the same methodological approach used in Vulcano, i.e. in situ NPP estimates through O 2 determination in the same incubation chambers used in Vulcano.

Plant incubations.
Cymodocea nodosa plants, each unit consisting of one rhizome with 6-7 shoots and corresponding roots, and water were collected at each donor site (CO 2 and Reference) and quickly transported (less than 5 minutes) to an intermediate incubation location, halfway of donor sites, where all incubations took place at 2 m depth. Plants were gently cleaned from sediment and epiphytes and placed into the incubation chambers with water from their own collection site (in situ incubations) or from the other site (transplant incubations).
For each type of incubation (in situ or transplants) two runs were conducted, one at midday (around 12:00 h) and one in the afternoon (around 16:00 h). At each run, four incubation chambers from each site were incubated together and thus they were exposed to the same irradiance. Light measurements (Licor sensor) were done at the incubation location along the incubation time. No significant differences were found between the midday (t-test, p = 0.7) or the afternoon (p = 0.2) irradiances during the two consecutive days of incubations.
Incubation chambers consisted of a gas-tight polyethylene plastic bag (30 cm long, 20 cm wide, 2.5 L approx.) with one sampling port to withdraw water samples (Fig. 5b). Chambers were sealed with a clamp made of a polyvinyl chloride (PVC) half-cylinder adjustable to a PVC rigid bar (Fig. 5a). Chambers were mounted and fixed in a steel bar and incubated together at 2 m depth (Fig. 5c). The positive buoyancy and flexible nature of the plastic bags allowed propagation of external turbulence to the interior of the chamber to prevent too thick surface Scientific RepoRts | 7:42278 | DOI: 10.1038/srep42278 boundary layers 55 . Incubation time ranged between 1-2 h to avoid within chamber O 2 saturation effects 56 . Water samples were taken at the beginning and end of each incubation and rapidly preserved for oxygen determination (see below). Plants were weighted and measured for NPP normalization and a shoot subsample was collected for gene expression analysis (see below).
In Ria Formosa, in situ incubations (n = 64) were conducted at 2 m depth to measure the productivity of C. nodosa to varying CO 2 concentrations. Seawater used in the incubations was collected from the field and bubbled with air at different concentrations of CO 2 , ranging from ~250 up to ~1500 μ atm (pCO 2 ) (i.e. ~2000 to ~2400 μ mol Kg SW −1 DIC and 8.39 to 7.70 pH (NBS scale)), thus covering the range found in Vulcano and the predicted CO 2 scenarios for the end of the century 1 without modifying any other seawater parameters. Water and biomass samples were processed as in Vulcano.
Gene expression. Material collection, RNA extraction and cDNA preparation. One shoot from three randomly chosen incubation chambers was collected at the end of both in situ and transplants incubations. Apical (youngest) shoots were avoided and only middle parts of the youngest fully mature leaves (2° or 3° rank leaves) of intermediate shoots were selected to standardize leaf developmental stages. Immediately after collection, leaves were blotted dry with tissue paper and preserved in RNAlater© (Ambion, Life technologies) to protect RNA from degradation. The whole procedure, from collection to fixing, lasted less than 5 minutes. After one night at 4 °C, samples were stored at − 20 °C until RNA extraction. RNA was extracted from C. nodosa leaves with the Aurum ™ Total RNA Mini Kit (BIO-RAD) following manufacturer's instructions. The full protocol for RNA extraction and cDNA preparation is described in Appendix 2, Supplementary Information.
Gene selection and RT-qPCR. Reverse Transcription-quantitative Polymerase Chain Reaction (RT-qPCR) analysis was performed to investigate differences in expression levels of genes belonging to major pathways affected by changes in CO 2 availability (i.e. photosynthesis, carbon assimilation pathways) that can be directly correlated with plant productivity (psaJ, psaC, psbA, psbD, LHCA1, FD, rbcL, ATPA, PEPC, SUS and BCA). Additionally, some studies have related high-CO 2 environments with antioxidant and oxidative stress responses in land plants, and more recently in other seagrass species growing close to natural volcanic CO 2 vents 22 , so a number of genes involved in free-radical detoxification and oxidative-stress response were also targeted (SOD, CAT, APX7, APX6, LBP, GSH-S and GR). Finally, since the presence of metals and toxic trace elements has been shown in fluid emissions near some vents, also two genes related to metal detoxification (MTP and MT) were selected. Specific primer pairs were established and tested for 8 putative reference genes (RGs) and 20 genes of interest (GOIs). (The full RT-qPCR protocol and gene description for selected genes are available in Appendix 2 and  Supplementary Table S10. Net plant productivity. Net plant productivity (NPP) was estimated following the methodology described in ref. 56. Initial and final water samples were collected from each chamber with plastic syringes (100 mL) at the beginning and end of the incubations, respectively. Once collected, syringes were quickly transported to shore for sample fixation (2 m away from the incubation site). For oxygen fixation, Exetainer vials (Labco Limited) (12 mL) (n = 3) were filled with water from each syringe. Dissolved oxygen concentration (O 2 ) was determined spectrophotometrically by the modified Winkler method according to the protocol described in ref. 57. Net plant productivity (NPP, μ mol O 2 g DW −1 h −1 ) rates were calculated by the difference between final and initial oxygen concentration normalized by the incubation time, the volume of water and the biomass incubated.
Statistics. Normality (Shapiro) and homoscedasticity (Levene) tests were run to test parametric assumptions.
Differences in seawater parameters between sites in Vulcano were tested with paired t-tests.
For gene expression, multivariate differences in gene expression among different treatments were tested by an analysis of similarity (ANOSIM), with Primer 6 (v.6.1.12). To inspect the significant gene regulation in the different conditions (in situ incubations and transplants) relative to controls, we used the method developed in REST 2009 Software version 2.0.13 58 . REST 2009 Software provide a means for determining the mean output and a p value for the likelihood of upregulation or downregulation using a hypothesis test. Statistical significance was set at p < 0.05.
To investigate the effect of CO 2 on productivity (NPP) of C. nodosa plants long-(in situ incubations) and short-term (transplants incubations) exposed to high CO 2 water from CO 2 vents, an ANCOVA analysis considering two factors, the "origin" of the plants (plants from CO 2 or Reference site) and the "water" used in the incubation (water from CO 2 or Reference site), and "irradiance" (as a continuous covariate) was performed in order to account for any potential effect of irradiance on NPP. ANCOVA analysis indicated a highly significant main effect of the "water" (p = 0.002) whereas the effects of "irradiance" or "origin" were not significant. A significant three-way (irradiance, origin, water) interaction was also revealed (p = 0.03), but the post-hoc tests did not provide a clear interpretation of the occurrence of any other relevant processes, i.e. it was not clear what explanatory factors needed to be retained in the model. We then tested for a model simplification using an Akaike's information criterion (AIC) test among 3-way and 2-way factorial and additive possible ANCOVA and ANOVA models. The 2-way ANOVA (two fixed crossed factors) with "origin" and "water" as explanatory factors produced the lowest AIC value, that is, the best fit and explanatory power indicating that a simplification of the ANCOVA model was justified 59 . We retained the two-way ANOVA (two fixed crossed factors) as the most adequate model to interpret the NPP results. Vulcano NPP data were log-transformed to meet parametric assumptions.
Likewise, at Ria Formosa the incubations were carried out along the day with changing irradiances. A multiple regression analysis including irradiance and DIC as explanatory variables was performed to investigate the effect of CO 2 on productivity (NPP) of C. nodosa plants. The explanatory power of DIC was highly significant (p = 0.01) as opposed to irradiance (p = 0.2) and the best AIC fit was obtained with DIC as single explanatory variable. For this reason we performed a linear regression with DIC as single explanatory variable.
To analyse the combined response of gene expression patterns and NPP on plants from in situ and transplant incubations, a Principal Component Analysis (PCA) and a Cluster analysis were also performed with the software PAST (v.3.03). To link gene expression to productivity a Spearman's Rank correlation analysis was conducted between the NPP and expression of each gene measured (n = 20 genes) considering the mean gene expression of each incubation type (n = 4 ranks, i.e. in situ and transplant incubations for both Reference and CO 2 plants). Unless otherwise indicated, data were analysed using R 60 .