Relationships between key functional traits of the waterlily Nuphar lutea and wetland nutrient content

Studied wetlands

The study focused on 11 populations of N. lutea collected in 11 wetlands dispatched along three rivers in the eastern part of France: the Ain, the Doubs and the Loue Rivers, all three of which belong to the Rhône watershed and flow on the calcareous Jura massif (Fig. 1). The wetlands were selected to maximise habitat contrast (mainly for nutrient loads and temperature) and because populations of N. lutea were sufficiently abundant, so that the sampling effort had a negligible impact on the population size at the wetland scale. In Doubs and Loue wetlands, N. lutea stands were rather monospecific, whereas in the Ain wetlands, N. lutea stands were surrounded by other aquatic plant patches (Callitriche platycarpa, Berula erecta, Mentha aquatica, Sparganium emersum, Potamogeton natans, Groenlandia densa, Myriophillum verticillatum).

Physicochemical characteristics of water

To assess whether the habitat conditions in wetlands affect plant traits, we documented monthly over a year cycle, from April 2015 to March 2016, the water parameters that may directly (nutrients, temperature) and indirectly (dissolved oxygen, phytoplankton abundance, dissolved organic carbon (DOC)) influence N. lutea traits. The physicochemical dataset, therefore, resulted in nine physicochemical parameters, measured once a month for 12 months, in the 11 studied wetlands.

The water temperature was measured automatically using submerged probes (HOBO UTBI-001 Tidbit; Onset Computer Corporation, Bourne, MA, USA) that registered the water temperature each day at 12:00 AM (UTC). The probes were fixed at a few metres from the 11 N. lutea populations, leaving them accessible for monthly surveys without disrupting the development of the studied populations (i.e. just outside the stand of N. lutea populations).

The other physicochemical parameters were recorded each month in each wetland near the temperature probes. We measured the electrical conductivity in situ at a 10-cm depth and the pH using a conductivity WTW sensor (TetraCon^® 325; Onset Computer Corporation, Bourne, MA, USA) and a pH WTW sensor (SenTix^® 41; Onset Computer Corporation, Bourne, MA, USA), respectively. Dissolved oxygen in water is ruled by photosynthetic activity, water motion and oxygen demand for organic matter mineralisation. Consequently, we documented monthly oxygen (WTW sensor, CellOx^® 325-3; Onset Computer Corporation, Bourne, MA, USA), DOC and chlorophyll-a concentrations in water.

Concentrations in nitrate and ammonium nitrogen, phosphate and DOC of the water were measured in the laboratory. For this purpose, at each sampling date, 500 ml of water was collected, brought back to the laboratory in a cooler, filtered the day of collection (polycarbonate filters with a mesh size of 0.2 µm), stored at 4 °C, and analysed within 48 h. The P-phosphate [P-PO₄³⁻] concentration was measured using the molybdenum blue colouration method after acidification (Rodier et al., 1996). The N-ammonium [N-NH4⁺] concentration was measured with the indophenol blue method (Berthelot’s reaction, Merck Spectroquant 14752). The N-nitrate [N-NO3⁻] concentration was measured using a Dionex ICS-1000 ion chromatograph. The DOC concentration was measured with a TOC analyser (Vario TOC Cube; Elementar, Langenselbold, Germany). To assess the chlorophyll-a concentrations, three liters of water were sampled and filtered through a 200-µm filter to remove zooplankton and any detritus and further filtered through 0.7-µm mesh filters, which were stored at −20 °C until extraction (overnight at 4 °C in acetone (90%)). Chlorophyll-a concentrations were measured the next day using a spectrophotometer (calculation taking into account measures done with λ = 630, 645 and 663 nm; Yentsch & Menzel, 1963). Dissolved oxygen in water was measured in situ 10 cm below the water surface for each sampling date using a WTW oxygen sensor.

The averaged values obtained for each parameter are grouped in Table S1.

Nutrient loads of wetland sediment

We measured nitrogen, phosphorus and organic carbon sediment concentrations in the 11 wetlands. For this purpose, we sampled sediment using a nine cm-diameter sediment corer just outside the stands of N. lutea in November and December 2014. The nitrogen and phosphorus concentrations were analysed (i) in the upper fraction of two centimeters of the sediment core because of the high bioavailability of nutrients and the high nutrient exchanges between water and sediment in this layer, and (ii) in the 0–40 cm fraction of the sediment core to take into account all the nutrients accessible by the roots of N. lutea. Each sediment sample was measured for total phosphorus (inductively coupled plasma—atomic emission spectrometry, iCAP6500 Radial; Thermo Fisher, Waltham, MA, USA), nitrogen (thermic catalytic oxidation, thermic conductivity detection, macro analyser Variomex; Elementar), and organic carbon concentration (thermic catalytic oxidation, infra-red detection, Vario TOC Cube; Elementar, Langenselbold, Germany).

The averaged values obtained for each nutrient are grouped in Table S2.

Plant traits sampling and analyses

Traits related to four sets of traits (growth, reproduction, defence and storage) were sampled (Table 1). Concerning growth, plant size is frequently used as a reference trait for terrestrial plants. Despite their strong indicative value, traits related to plant size are rather difficult to quantify for aquatic plants, and more specifically, for N. lutea. Indeed, plant height is controlled by water depth. Moreover, the individual size is frequently impossible to assess because of the very large size of N. lutea clones due to its extensive rhizomes, which are usually several metres long, branched in all directions, intermingled and strongly and deeply anchored in the underwater sediment. Furthermore, we wanted to relate plant traits to the environmental strains recently encountered by the plant. For this reason, we selected plant traits independent from water depth and individual long-term history and focused on traits related to recent growth of the plant, informing on the conditions it encountered during or just before habitat sampling. The rhizome length produced per apex per year and the number of leaves produced per apex per year were consequently selected to qualify the growth of N. lutea and documented on the part of the rhizomes corresponding to the last growing year. For reproduction, we documented the number of flowers produced per apex per year (counting the flower peduncle scars in rhizomes, as described later in the ‘Rhizome traits’ section), the number of seeds produced per fruit, the fruit water content and the seed:fruit mass ratio. For defence, we focused on floating leaf blades (which face high grazing pressure, Stenberg & Stenberg, 2012; Klok & Van Der Velde, 2019), because submerged leaves were absent of many sites, particularly the eutrophic ones (despite the fact that submersed leaves may be more consumed by herbivores as ducks and swans, Cronin, Wissing & Lodge, 1998). For this purpose, the leaf water content (i.e. the floating leaf blades water content) and the C:N ratio, related to the tissue cost and induced resistance (nitrogen enrichment favour the production of low-density tissues, and reduced nitrogen availability induces the production of high-density tissues rich in lignin and structural carbohydrates, Garnier & Laurent, 1994; Poorter & De Jong, 1999; Craine et al., 2001) were measured. Sclereid abundance was also measured because it is considered, as high density tissues, to prevent grazing and increase plant rigidity (Konno, Inoue & Nakamura, 2014). For storage, macrophytes accumulate starch in their rhizomes (Kubin, Melzer & Čížková, 1994). We measured the rhizome starch and water contents. Starch accumulation in rhizomes potentially leads to the increase of rhizome diameter, and so, the diameter and the volume of the rhizomes were also measured.

Plant samples were collected in the field on June 23, 2015 for the Doubs and the Loue Rivers and on June 24, 2015 for the Ain River. For each wetland, we cut the end of five growing rhizomes (approximately 80 cm length, including apex and the following rhizome), five fully developed undamaged floating leaf blades and 10 full-grown fruits were randomly sampled in the population (Table 1). The number of plants sampled per wetland was a compromise between (1) the duration of the data sampling for a given population (trait documentation being time-consuming and having to be documented rapidly after the sampling), (2) the necessity to collect rhizomes being at the same growth stage in the sampling season and (3) the fact that sampled wetlands were located far from each other. The plants were collected on a circle whose centre was the water sampling point. Each plant sample was collected at a minimum distance of 10 m from each other. When the population was discontinuous, a sample was collected in each plant patch (VILC and VILM wetlands). Rhizome fragments were collected in order to obtain the apex and at least 3 years of growth. The first fully developed and mature floating leaf blades from the centre of the apex were sampled (which may reduce strongly the developmental effect on sclereid abundance; see below). For the wetlands of the Ain River, only three and five fruits were sampled for VILC and SBR, respectively, because of the strong herbivory pressure on flowers and fruits (many peduncles were found without fruits, possibly consumed by ducks or swans). The N. lutea traits dataset, therefore, results in 13 traits, randomly measured once in five or 10 replicates (depending on the trait; see above), in the 11 studied wetlands.

Statistical analyses

The normality and homoscedasticity of the data were tested using the Shapiro–Wilk, Levene and Bartlett tests. When necessary (rhizome volume), the data were log-transformed to reach normality. The number of flowers per year, which followed a Poisson model, was not included in linear models (LM) (see below).

A principal component analysis (PCA) was first performed to discriminate wetlands according to the physicochemical characteristics of the water. The median values of the 12 measures of each of the water physicochemical characteristics (1/month) were considered in the PCA. The scores of each water physicochemical characteristic on the PCA axes were verified using bootstrapping. Bootstraps were replicated 10,000 times, and bootstrap confidence intervals of 95% were constructed. Then, to identify the potential role of the wetland physicochemical characteristics on N. lutea traits, we performed LM. LM related traits (except the number of flowers) with (i) the coordinates of the 11 wetlands along the two first axes of the PCA of the physicochemical characteristics of water, and with (ii) the nitrogen and phosphorus sediment concentrations (the organic carbon was removed from this analysis because it was highly correlated with the sediment nitrogen, R² = 0.87 for the upper two centimeters and R² = 0.77 for the upper 40 cm). Because they are counts, the number of flowers, leaves and seeds were analysed with generalised linear models (GLM). Because the variance of the seed number is much larger than its mean, a negative binomial GLM was used for this trait. The number of leaves and seeds was therefore analysed with both LM and GLM.

Some authors proposed that the ecosystems may change from nitrogen-limited functioning to phosphorus-limited functioning depending on the value of the N:P ratio in plant tissues (N/P < 18 and N/P > 24, Suttle & Harrison, 1988; N/P = 20, Güsewell, Koerselman & Verhoeven, 2003). We tested whether this trophic partitioning of ecosystems may lead to significant differences in plant traits. These thresholds led to the same wetland partitioning: four wetlands (GRI, CHA, LON, CHEM) had N:P ratios <20 and were considered as nitrogen-limited (i.e. led by the phosphorus), while the seven others had N:P ratios >30 and were considered as phosphorus-limited (i.e. led by the nitrogen). Some boxplots were performed to determine the potential consequences of the ecosystem trophic functioning on traits of N. lutea. The significance of the difference between the two groups was tested thanks to a t-test for each trait.

The ‘Holm sequentially rejective Bonferroni test’ (Holm, 1979) was used to adjust the p-values of all tests and to counteract the problem of multiple comparisons. It was used because it is considered to be more powerful than the Bonferroni correction (Holm, 1979; Abdi, 2010). However, because the Holm adjustment inflates the type 1 error (Abdi, 2010), we decided to keep both the original and corrected p-values.

The α-value was set to 0.05.

All the analyses were done using RStudio (Version 1.0.153) and the R package (R Core Team, 2017).

Physicochemical characteristics of wetlands

The first and second axes of the PCA (Fig. 2) explained 45.5% and 29.3% of the variance of the physicochemical dataset, respectively (Table S1). All scores of each water physicochemical characteristic on the PCA axes were within the bootstrap confidence intervals of 95%. The first axis (PC1) was mainly positively correlated with phosphates, chlorophyll-a and organic carbon concentrations in water and negatively with the nitrate concentration and to a lesser extent, to conductivity (Fig. 2A). Both Ain and Loue wetlands (CLO, VILC, CDC and SBR) were plotted on the negative part of the axis, whereas three Doubs wetlands (CHA, CHEM and LON) were plotted on the positive part of the axis (Fig. 2B). The PCA separated wetlands according to the nutrient concentration in water (PC1), and negative values were related to wetlands with higher nitrate concentrations (0.008 mg.l⁻¹< [N-NH₄] < 0.027 mg.l⁻¹; 1.11 mg.l⁻¹< [N-NO₃] < 4.99 mg.l⁻¹; 0.006 mg.l⁻¹< [P-PO₄] <0.014 mg.l⁻¹ for CLO, VILC, CDC and SBR), while positive values were correlated with wetlands having higher phosphate and ammonium concentrations (0.020 mg.l⁻¹ < [N-NH₄] < 0.039 mg.l⁻¹; 0.80 mg.l⁻¹ < [N-NO₃] < 1.17 mg.l⁻¹; 0.046 mg.l⁻¹ < [P-PO₄] < 0.087 mg.l⁻¹ for CHA, CHEM and LON) (Table S1). The second axis (PC2) was negatively correlated with temperature, ammonium concentration and conductivity and to a lesser extent, with the pH and dissolved oxygen (Fig. 2A). Most wetlands were plotted on the positive part of this axis, but two large oxbow lakes (CHA and MER, which had higher average temperatures of 15.1 °C and 14.5 °C, respectively) and ammonium nitrogen concentrations in water (0.038 and 0.039 mg.l⁻¹, respectively) were plotted on the negative side of the axis (Fig. 2B). No general grouping by river was visible.

Concerning the sediment nutrient concentrations, three wetlands (MER, CDC and CLO) had high nitrogen and organic carbon concentrations (positively correlated between each other, R² = 0.87 for the upper two centimeters and R² = 0.77 for the upper 40 cm). The phosphorus sediment concentration was higher in CHA compared to the other wetlands. These results were the same for the two considered sediment layers (both 0–2 and 0–40 cm layers).

Relationships between plant traits and physicochemistry of wetlands

LM and GLM results were congruent between each other considering traits analysed by the two modelling methods (leaf and seed numbers, Table 2).

Concerning the physicochemical characteristics of water, only the rhizome length was positively correlated to wetland coordinates along PC1 and PC2 among the growth traits (Table 2). Among reproduction traits, the number of flowers and seeds and the seed:fruit mass ratio were positively correlated to the wetland coordinates along PC1. The seed:fruit mass ratio was also negatively correlated to PC2 (Table 2). Among storage traits, the diameter and the volume of rhizomes were positively correlated to wetland coordinates along PC1 (Table 2). The rhizome water content was negatively correlated with the wetland coordinates along PC2. For defence traits, the sclereid relative abundance was negatively correlated and the leaf water content was positively correlated to the wetland coordinates along the second axis of the PCA, and the C:N ratio was correlated to both PC1 and PC2 (positive correlation on PC1 and negative correlation on PC2; Table 2).

After the adjustment of the p-values (Holm), all the traits but the rhizome length (on both PC1 and PC2), the rhizome water content (on PC2) and the sclereid relative abundance (on PC2) remained significantly related to water characteristics.

Concerning the sediment nutrient concentrations, the results were almost identical depending on the layer considered (both 0–2 and 0–40 cm layers; Table 2). The rhizome length was positively correlated with the phosphorus sediment concentration in both sediment layers, and negatively correlated with the nitrogen concentration in the upper two centimeters of sediment (Table 2). For the reproduction traits, the seed:fruit mass ratio was positively correlated with both the nitrogen and organic carbon sediment concentrations (upper two centimeters of sediment), and with the phosphorus sediment concentration (both sediment layers). The numbers of flowers and seeds were negatively correlated with both the nitrogen and organic carbon concentrations in both sediment layers (Table 2). The numbers of seeds was also positively correlated with the phosphorus concentration when considering the 0–40 cm sediment layers (Table 2). The rhizome diameter was negatively correlated with both the nitrogen and organic carbon sediment concentrations in both sediment layers (Table 2). The rhizome volume was negatively correlated with both the nitrogen and organic carbon in the upper two centimeters of sediment, and positively correlated with the phosphorus in the upper 40 cm of sediment (Table 2). Concerning the defence traits, only C/N was positively correlated with the phosphorus concentration in both sediment layers (Table 2).

After the Holm adjustment of the p-values, only the seed:fruit mass ratio remained correlated with phosphorus in the upper two centimeters of wetland sediments.

Trait contrast according to the nitrogen or phosphorus trophic limitation of wetlands

Among the growth traits, only the rhizome length differed between the nitrogen- and the phosphorus-limited wetlands (p-value = 0.018), and it reached higher values for the nitrogen-limited wetlands (Fig. 3). For reproduction traits, the number of flowers was higher for the nitrogen-limited wetlands (p-value = 0.012, Fig. 3). Two other reproduction traits (the water content of the fruits and the seed:fruit mass ratio) tended to be higher for nitrogen-limited wetlands, but without reaching significance: the water content of the fruits and the seed:fruit mass ratio (p-value = 0.058 and 0.059, respectively). Among the storage traits, both the rhizome volume and starch content differed among the nitrogen- and phosphorus-limited wetlands (p-value = 0.027 and 0.011, respectively). The rhizome volume was higher in the nitrogen-limited wetlands, while the starch content was higher in the phosphorus-limited wetlands (Fig. 3). None of the defence traits varied significantly between the nitrogen- and phosphorus-limited wetlands.

After the Holm adjustment of the p-values, no relationship remained significant.

Relationships between wetland physicochemical characteristics and N. lutea traits

Water resource conditions

The PC1 of the PCA outlined the functional contrast of the resource conditions, and particularly between the phosphorus-rich (positive values along the axis) and the nitrogen-rich (negative values along the axis) aquatic ecosystems. The 11 wetlands were arranged along this resource gradient. The factorial coordinates of wetlands along this axis were strongly correlated to their trophic state index (Carlson, 1977; Table 3).

Table 3:

Trophic state index (TSI) of Carlson (1977) applied to the 11 studied wetlands, calculated with the chlorophyll-a (Chla) average concentration and the total phosphorus average concentration (TP) of the water.

The Trophic State Index of Carlson varies between zero and 100, zero characterising highly oligotrophic water and 100 characterising highly eutrophic water. The significance and location of the wetland codes are indicated in Fig. 1.

Floodplain	Wetland	TSI (Chla)	TSI (TP)	MOY
Ain	SBR	49.1	30.7	39.9
	VILC	44.4	35.3	39.8
Doubs	BING	54.4	57.7	56.0
	CHA	61.3	64.5	62.9
	GRI	53.9	64.7	59.3
	LON	60.3	68.6	64.4
	MER	55.9	37.4	46.6
Loue	BAR	62.1	60.5	61.3
	CHEM	62.3	59.4	60.8
	CDC	41.0	40.2	40.6
	CLO	48.0	42.5	45.3

DOI: 10.7717/peerj.7861/table-3

The four sets of traits responded differently according to the resource conditions (Fig. 4). Concerning growth related traits, the rhizome length was significantly and positively correlated with PC1, that is, the resource conditions. Both rhizome diameter and volume were considered storage traits in our study. Unexpectedly, they increased with the phosphate concentration in water, while storage is usually considered to increase with habitat adversity (Grime, 2006; Grasset et al., 2015). The rhizome volume was significantly higher in phosphorus-rich wetlands and did not correlate with the rhizome starch content (that was higher in phosphorus-poor wetlands). This suggests that diameter and volume may be considered as growth instead of storage traits. The increase in water phosphate concentrations was therefore related to the increase in rhizome growth traits (volume, diameter and length).

Some other growth traits, such as leaf length and leaf area index have previously been demonstrated to increase with water nutrient content and particularly with phosphates (Klok & Van Der Velde, 2017). N. lutea is considered to be a top-competitor in standing waters (Arts et al., 1990; Bornette et al., 2008; Bornette & Puijalon, 2011; Grasset et al., 2015). As size traits are frequently related to plant competitive ability (Grime, 2006), decreasing water and sediment nutrient load (especially the phosphate concentration) may decrease N. lutea competitive ability.

The number of flowers and seeds were also significantly and positively correlated with the increase in phosphorus and organic carbon in water, demonstrating a positive relationship between these nutrients and the reproduction traits.

Concerning the defence traits, the C:N ratio was significantly and unexpectedly positively correlated with the increase in phosphorus and organic carbon in water. According to the conservation strategy, leaf tissue resistance usually increases when the resource conditions of the habitat decrease (Ordoñez et al., 2009). The observed increase in the C/N ratio may reflect the consequence of an increasing plant size with the increase in phosphorus and organic carbon in water, leading to a higher number or size of leaves and petioles, which requires structurally resistant tissues. Even if starch storage may relate to reserve accumulation for spring sprouting, the present results suggest that it is also related to stress-tolerance (low phosphorus concentrations in water), as structural carbohydrates, which are frequently outlined in the literature (Gebauer, Reynolds & Tenhunen, 1995). This suggests the occurrence of different storage tactics according to plant strategy and/or phenology. Competitive species, such as N. lutea, may store starch for both seasonal regeneration and stress tolerance, whereas stress-tolerant species may also store structural carbohydrates for increasing tissue lifespan (Grasset et al., 2015). Finally, N. lutea, a late successional species, may be more plastic in its morphological variables than in its metabolic variables.

Consequences for the ecological niche of N. lutea

The variation in N. lutea growth and reproduction traits was mostly affected by phosphorus availability, and nitrogen seemed to play a minor role in its phenotypic variations. However, nitrogen was always abundant in the 11 studied wetlands, making it difficult to efficiently measure nitrogen-limitation. Although nitrogen is a secondary nutrient in freshwater ecosystems in comparison to phosphorus (Suttle & Harrison, 1988; Güsewell, Koerselman & Verhoeven, 2003), some studies highlighted alterations in plant traits due to a lack of nitrogen in aquatic ecosystems (Feller et al., 2003).

Nuphar lutea is a long-living clonal species that may face huge contrasts in environmental conditions through its life. In such a situation, phenotypic adjustments may be essential for facing environmental variations. Our results highlighted that the performance-related traits of N. lutea seemed to be improved by phosphate and/or organic carbon in water, particularly for both growth and reproductive success, at a short time scale (over a year). On the contrary, the performance of N. lutea decreases when it faces nutrient stress: its growth and reproductive success seem to be forsaken by the increase of defence traits. The absence of variation in some traits, such as leaf number and fruit water content, which appear unrelated to the measured environmental characteristics, suggests either that the variation in such traits has no consequence for N. lutea or that these traits are constrained by N. lutea morphology, or that their variation is constrained by environmental variation that occur at larger temporal or spatial scales. For example, N. lutea only produces one flower per spiral of leaves each year, and the number of flowers may only increase with the increase in spiral number, that is, with a strong increase in rhizome diameter, which only occurs through over several years of growth. In the same way, the surprising absence of correlation between the rhizome starch content and habitat severity (i.e. high ammonium concentrations and low oxygen concentrations) suggests that the species is unable to store starch in the case of anoxic stress, or that this stress may not be mitigated by reserve accumulation. Either the species is unable to tolerate this stress, or it uses some other strategy (e.g. accumulation of structural carbohydrates in rhizomes, for example, which may explain the decrease of rhizome water content with habitat severity).

Nuphar lutea is a top-competitor species (Bornette et al., 2008), generally favoured in nutrient-rich habitats. However, such nutrient-rich habitats may disfavour the species if anoxia becomes so high that the species is unable to counterbalance its negative effect through its oxygenation of the rhizosphere. Global warming and eutrophication should consequently lead to N. lutea progression as long as temperature and productivity does not exceed a threshold value beyond which the plant will be unable to maintain a sufficient degree of oxidation of the rhizosphere for its survival.

Adverse conditions may also occur in highly oligotrophic conditions, such as in cold phosphorus-poor wetlands.

Our study aimed as searching the relationships between plant traits and environmental variables in a large set of wetlands over a year study. The fact that N. lutea is a long-lived plant species may lead to environment-species traits relationships that occur at a longer time scale than the one year of this study. In the same way, the very large size of rhizomes may lead to an integration of the response to the environmental signal at a very large scale, leading to a possible attenuation of the response measured at the rhizome fragment scale, even if previous studies outlined the reactivity of plant ramets to habitat variability (Hutchings & Price, 1993). This may be true for sediment characteristics, as they may vary strongly both horizontally and vertically.

Even if the environmental characteristics of wetlands were rather contrasting, we may have underestimated the effect of environmental characteristics on plant traits. For example, we did not sample habitats with a low pH, but such habitats may have induced different plant responses (e.g. because of low carbon availability, low production and low mineralisation rate of organic matter, Kok, Van Der Velde & Landsbergen, 1990). In the same way, the temporal scale of the study did not allow to assess the effect of environmental variations that occur over longer time scale. For example, river incision, leading to the oligotrophication and cooling of wetlands along karstic rivers (Bravard et al., 1997), may affect adversely N. lutea at a time scale larger than one year, decreasing progressively its health through time.