To test the effect of temperature and sediment nutrient content on Elodea nuttallii, a full-factorial microcosm experiment was set up. Shoots of E. nuttallii were collected from a small pond on the grounds of The Netherlands Institute of Ecology (NIOO-KNAW), Wageningen, The Netherlands (51°59′15.0″N; 5°40′14.8″E) on 07-09-2015. After collection, the plants were rinsed and acclimatized at room temperature for 2 days prior to the start of the experiment.

The experiment was carried out in 4 L plastic microcosms (14 × 14 × 21 cm), which contained 1.1 L of sediment and 2.7 L of water. Nutrient treatments were achieved by mixing artificial pond sediment (20% organic matter, Velda, Enschede, The Netherlands) with sand and consisted of 12.5, 25, 50, and 100% (v/v) of pond sediment (n = 5), covered with a one centimeter layer of sand. The artificial pond sediment contained 31 ± 1.8, 0.80 ± 0.048, and 0.11 ± 0.0084% (mean ± SE) C, N and P respectively. One shoot fragment of E. nuttallii of ± 5.5 cm (C:N = 19 ± 1.4, C:P = 435 ± 72; mol:mol; n = 5) was placed in the middle of each microcosm and the microcosm was topped off with nutrient-poor tap water [3.5 ± 0.5 (mean ± SE) μM DIN and undetectable levels of DIP]. The microcosms were placed in four aquaria, which served as temperature-regulated water baths. The temperature treatments were 10, 15, 20, and 25°C, which were obtained by a computer-controlled (Specview 32/859, SpecView Ltd., Uckfield, UK) custom-made climate control system. These temperatures are within the range of natural temperatures E. nuttallii would encounter, as water temperatures in the Netherlands vary seasonally between 4 and 23°C (van Dam, 2009). Light (14:10 hours light:dark) was provided by two 28W TL5 HE lamps (Philips, Eindhoven, The Netherlands), hung above the aquaria with an average light intensity at the water surface of 30 μmol s⁻¹ m⁻². To ensure equal light conditions between treatments, position of the microcosms in the water bath was randomized once a week and evaporation losses were compensated by additions of demi-water. To prevent excessive periphyton and phytoplankton growth during the experiment, one periphyton-grazing snail (Planorbarius corneus) and one filtering mussel (Dreissena polymorpha) were put in each microcosm. In pilot tests, P. corneus did not feed on E. nuttallii (Peiyu Zhang, personal observation) and no grazing on the plants was observed during the experiment. The snails were retrieved from the same pond as the plants, and the mussels were collected from the Nether Rhine, Wageningen, the Netherlands (51°57′12.9″N 5°39′48.2″E). In case either snail or mussel died, another one was added.

After 58 days the experiment was terminated. From the middle of each microcosm, water samples were taken to determine dissolved nutrient concentrations, filtered over prewashed GF/F filters (Whatman, Maidstone, U.K.) and stored at −20°C until further analysis. Samples for pore water nutrients were taken in each microcosm through a 10 cm Rhizon SMS (Rhizosphere, Wageningen, the Netherlands) and stored at −20°C until further analysis. Plants were cut at the sediment level, and the above- and belowground biomass was harvested and rinsed with demi-water. All plant materials (above- and belowground) were dried at 60°C until constant dry mass and weighed. During the harvest, basic parameters were measured that describe the growing conditions (pH, alkalinity and seston chlorophyll-a). Methods and results of these measurements can be found in Supplementary Material S1.

Plant material was grinded to a fine powder on a microfine grinder (MF 10 basic, IKA-werke, Staufen, Germany) or in test tube with a 1/8” ball bearing (Weldtite, Lincolnshire, UK) on a Tissuelyser II (QIAGEN, Germantown, USA). For nitrogen (N) and carbon (C) content, 0.2–2 mg dry mass was analyzed on a NC analyser (FLASH 2000 NC elemental analyser, Brechbueler Incorporated, Interscience B.V., Breda, The Netherlands). For phosphorus (P) content, 1–4 mg dry mass was combusted in a Pyrex glass tube at 550°C for 30 min. Subsequently, 5 mL of persulfate (2.5%) was added and samples were autoclaved for 30 min at 121°C. Digested P (as PO43−) was measured on a QuAAtro39 Auto-Analyzer (SEAL Analytical Ltd., Southampton, U.K.). Concentrations of dissolved nutrients (PO43−, NO2−, NO3− and NH4+) of thawed water-samples were determined on a QuAAtro39 Auto-Analyzer (SEAL Analytical Ltd., Southampton, U.K.). Results for the dissolved nutrients in the water column can be found in Figure S1.2.

Plant specific growth rate (SGR) was calculated with the following formula:

Where DW_t is the plant aboveground dry weight at the end of the experiment, DW₀ the dry weight at the beginning of the experiment (determined by multiplying the initial wet weight with the plants wet weight/dry weight ratio) and t the experimental duration (=58 days).

Data on above- and belowground parameters (specific growth rate, above- and belowground biomass, carbon:nutrient stoichiometry and elemental contents) and dissolved nutrient concentrations (DIN and DIP) in the water column and the pore water were tested for effects of temperature, nutrients and their interaction with generalized linear models (function glm from stats package). Visual examination of the data distribution (function hist) led to the use of a gamma distribution. Post-hoc tests within treatment levels were carried out using Tukey contrasts [function glht from multcomp package (Hothorn et al., 2008)], with P-values corrected for multiple comparison as described by Benjamini and Hochberg (1995).

A systematic literature review was carried out in Web of Science based on the guidelines described by the Collaboration for Environmental Evidence (2013). The search term (“submerged macrophyte^*” OR “aquatic plant” OR isoetid OR macrophyte^* OR “aquatic weed” OR seagrass^*) AND (stoichiometr^* OR “^*chemical composition” OR “nutritional quality” OR “nutrient composition” OR “elemental composition” OR “nutrient content” OR “nutrient ratio^*” OR C:N OR C:P OR N:P OR “plant nutrient concentration^*”) AND (warming OR eutrophication OR temperature^* or enrichment or fertilis^* or “nutrient availability”) on 01-11-2016 gave 414 hits. Further selection based on abstracts, graphs and tables led to 47 papers that contained information on temperature and/or nutrient effects on elemental composition of submerged angiosperms. Data originating from light limited conditions (as indicated in the paper itself) were excluded from analysis, as well as studies without reported standard errors or deviations, and studies with limited (n < 2) or non-reported sample size. From the selected papers, data on C:N and C:P ratios were extracted with use of Plotdigitizer and Engauge and converted to molar ratios when necessary. In addition, C, N and P contents, growth rates (above- or belowground), habitat (marine or freshwater), which part of the plant was analyzed (above- or belowground) and sample size were extracted when reported. If the described methodology indicated possible additional results that were not reported, corresponding authors were contacted to retrieve those data. If experiments reported several measurements over time, only the final measurement was extracted. If papers contained multiple experiments, on the same or on different species, these were extracted as being separate studies.

Control and elevated treatments were defined for both temperature and nutrient addition for each experiment separately. The lowest water temperature reported was defined as the control temperature treatment and 3–6°C above that temperature [equivalent to RCP scenario 8.5 from IPCC (2014)] was defined as the elevated temperature treatment. For the nutrient addition studies, those studies that manipulated both nitrogen and phosphorus simultaneously were selected. The lowest nutrient condition reported was defined as the control treatment and the highest as the elevated treatment. The data was then split up into above- and belowground plant responses, as different parts of plants were expected to respond differently (Bloom et al., 1985). These selection criteria led to a total of 50 studies on nutrient addition spread over 26 papers (of which 50 and 11 on above- and belowground responses respectively) and 3 studies on temperature (only on aboveground responses) originating from 3 papers. Temperature studies were all conducted in mesocosms, whereas nutrient studies included in situ fertilization experiments (38), mesocosm experiments (9) and laboratory experiments (3). Of the nutrient addition studies, 9 studies tested a range of nutrient concentrations of which the highest and lowest were selected, while the majority (41) specifically looked at the addition of nitrogen and phosphorus to the system relative to a control level. An overview of the dataset selection can be found in Figure S2 and an overview of the selected papers in Supplementary Material S3.

Delta response ratios and their variances were calculated for each separate study according to Lajeunesse (2015):

Where X denotes mean of the fixed factor of interest [C:N and C:P ratio, growth rate (μ) and C, N and P contents], SD the standard deviation of that mean and N the sample size.

All statistics were carried out in R (R Core Team, 2015). To test whether response ratios deviated from zero, mixed effect models were fitted to the response ratios and their variances with the function rma.mv [package metafor; Viechtbauer (2010)], incorporating reference and species as random effects. To test whether freshwater and marine systems differed in response ratio, separate models were compared for significant differences between the two habitat types (by adding habitat as a moderator to the function rma.mv).

Temperature affected the specific growth rate and above- and belowground biomass of E. nuttallii (Table 1). As indicated by the interaction term, temperature only affected specific growth rate at intermediate sediment nutrient content (e.g., 25%), with optimal growth at 15 and 20°C (P < 0.05, Tukey post-hoc comparison; Figure S4). Similarly, aboveground biomass was highest at these temperatures, irrespective of nutrient treatment (P < 0.05; Figure 1A). Belowground biomass of E. nuttallii was affected by temperature, and this effect interacted with nutrient treatment (Figure 1B, Table 1). The effects of temperature on belowground biomass seemed strongest in the lowest nutrient treatments (e.g., 12.5%), where biomass tended to increase 3-fold between 15 and 20°C but these effects were not significant in post-hoc tests (P = 0.09).

Sediment nutrient content affected the specific growth rate and the above- and belowground biomass of Elodea nuttallii (Table 1). However, no significant differences between nutrient treatments could be observed for specific growth rate (Figure S4), nor for aboveground biomass (Figure 1A) in the post-hoc comparisons. Belowground biomass decreased with increasing sediment nutrient content, and this effect interacted with temperature treatment (Table 1; Figure 1B). Belowground biomass tended to decrease 6-fold over the entire range of nutrient treatments at 25°C, but this effect was not significant (P = 0.08).

Temperature negatively affected aboveground C:N ratios (Table 1, Figure 2A), which was most visible at intermediate sediment nutrient content (25%). In this treatment, aboveground C:N ratios decreased moderately but significantly between 10 and 25°C (P < 0.001, Tukey post-hoc comparison). No effects of temperature on aboveground C:P ratios were observed, nor on belowground C:N and C:P ratios (Table 1, Figures 2B–D).

Sediment nutrient content affected aboveground C:P ratio, while no effect on aboveground C:N ratios was observed (Table 1). Aboveground C:P ratios of E. nuttallii were negatively affected by increasing sediment nutrient content (Figure 2B, Table 1). This effect was most visible at 15°C, where the C:P ratio significantly decreased 4-fold the entire range of nutrient treatments (P < 0.001, Tukey post-hoc comparison). Belowground C:N and C:P ratios were affected by nutrient treatment, and the effect on C:N interacted with temperature (Figure 2C). Belowground C:N ratios significantly increased between 25 and 100% nutrient treatments at 20°C, while belowground C:P ratios increased 4-fold between those nutrient treatments at the same temperature (P < 0.01; Figure 2D).

Accompanied by the changes in aboveground C:N ratio, temperature seemed to affect aboveground N content (Table 1). However, no differences between any of the temperature treatments could be detected in post-hoc comparisons (Figure S5B). Belowground carbon content was affected by temperature, and halved between 10 and 25°C in the lowest sediment nutrient treatments (12.5 and 25%; P < 0.05). No effects of temperature on aboveground C and P content were observed, nor on belowground N and P contents (Table 1).

Sediment nutrient content affected aboveground P content, with a 3-fold increase over the entire range of nutrient treatments (P < 0.05; Figure S5C). No effect on aboveground C or N content was observed for nutrient content (Figures S5A,B). Belowground C content significantly increased 14% over the entire range of nutrient treatments at 25°C (P < 0.01; Figure S5D), while belowground N and P content were not affected by nutrient treatment (Table 1).

Temperature affected dissolved nutrient concentrations in the pore water, and this effect interacted with nutrient treatment (Table 1). Temperature effects on pore water DIN concentrations were strongest at intermediate sediment nutrient content (50%), where values significantly doubled from 10 to 15°C, and decreased at higher temperatures (P < 0.01, Tukey post-hoc comparison; Figure 3A). Pore water DIP concentrations were significantly higher at 15°C than other temperature treatments in the highest nutrient treatment (P < 0.01; Figure 3B). This response was less pronounced in other nutrient treatments. Similarly to temperature, sediment nutrient content affected DIN and DIP concentrations in the pore water. Pore water DIN and DIP increased 7- and 16-fold, respectively, from the 12.5 to 50% nutrient treatment irrespective of temperature (P < 0.01; Figures 3A,B).

No significant effects of elevated temperature were observed on aboveground C:N and C:P ratios (Figure 4), nor on aboveground C, N and P contents (Figure S6A) or belowground N and P contents (Figure S6B). Sample sizes were too low to analyze effects of elevated temperature on aboveground growth rates (n = 0), on belowground C:N and C:P ratios and C content (n = 1) or on potential differences between marine and freshwater ecosystems (n = 1 and 2 respectively).

Nutrient addition significantly decreased aboveground carbon:nutrient ratios, with 24.7 and 21.9% for C:N and C:P ratios, respectively (Figure 5A). This decrease in aboveground carbon:nutrient ratios was accompanied by a 23.5% increase in aboveground N content and a tendency for increased P content (with 20.6%, P = 0.06), while C content remained unaffected (Figure S6C). Furthermore, aboveground growth rates tended to increase 83.1% with nutrient addition, but this effect was not significant (P = 0.08, Figure 5A). Similar to aboveground responses, belowground C:N ratio also declined 15.6% with nutrient addition, while no effect on belowground C:P ratios was observed (Figure 5B). This decline in C:N ratio was accompanied by an 18.2% increase in belowground N content (Figure S6D).

Aboveground carbon:nutrient stoichiometry of marine and freshwater plants responded qualitatively similar to nutrient addition, though the number of studies in the latter group was far lower (Figure 6). Quantitatively, responses in C:N and C:P were stronger for freshwater compared to marine plants (P < 0.001). Sample sizes were too low to analyze differences in aboveground growth rates between freshwater and marine plants (n = 0 for freshwater plants).