The experiments were conducted in 10 equally sized (ca. 0.08 ha) experimental ponds (N 30°17′17″, E 114°43′45″) being constructed from a lotus pond (culturing Nelumbo nucifera) by dredging surface sediments rich in nutrients and organic matter and introducing sediments and water from a nearby lake, Lake Bao’an (surface area 48 km², mean water depth, Z_M, and 1.9 m) (see Yu et al., 2015 for the initial environmental conditions of the 10 ponds). The experimental area is located in the middle Yangtze River Basin with a warm, humid subtropical climate (annual mean air temperature ca. 19°C and precipitation ca. 1030 mm). The monthly rainfall during the experiments was 125.7 mm in Yu et al. (2017) and 214.7 mm in this study (Supplementary Table 1). The mean air and water temperatures were 28.0 and 29.1°C in Yu et al. (2017) and 22.2 and 23.7°C in this study. Light intensity was 40998 and 38776 lux in this experiment and Yu et al. (2017), respectively. Mean Secchi depth (SD) in Yu et al. (2017) and this study were similar, being 52.1 and 51.9 cm, respectively.

A gradient of five target concentrations of total nitrogen (TN) (control, 2, 10, 20, and 100 mg L^-1) and two duplicates were made. The target TN concentrations of the control ponds (ca. 0.5 mg L^-1) functioned as background concentrations. In China, a nitrogen concentration of 2 mg L^-1 TN refers to V type of water based on environmental quality standards for surface water (General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China [AQSIQ], 2002b), 10 mg L^-1 NO₃-N refers to standards for drinking water quality (Standardization Administration of the People’s Republic of China [SAC], 2006), and 20 mg L^-1 TN refers to primary B based on the discharge standard of pollutants for municipal wastewater treatment plants (General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China [AQSIQ], 2002a). In part of the United States, 100 mg L^-1 NO₃-N refers to water quality standards for agriculture and livestock (Xia et al., 2004). Our gradient therefore covers the range observed in the real world. To maintain target concentrations, NH₄Cl fertilizer (NH₄Cl, ≥99.5%, Sinopharm Chemical Reagent Co., Ltd., Shanghai) was added every month relative to the difference between the measured and target concentrations. No phosphate fertilizer was added to the ponds in the experiment.

On 15 March 2013, P. crispus were collected from Lake Biandantang (N 30°32′, E 114°33′), a sub-area of Lake Bao’an. Similar-sized plants were selected and cut into a unified leaf length of 15 cm. The plants were then cultured in batches for 15 days in boxes (65 cm × 41 cm × 31 cm) filled with lake water. On 1 April 2013, similar-sized plants were selected and planted in plastic pots (23 cm in top diameter, 13 cm in bottom diameter, and 13 cm in height) filled with 10 cm sediments mixed with 1/3 washed sand (three plants in each pot). Sediment was taken from the experimental pond of the control treatment, with an original total nitrogen concentration of 1.93 mg g^-1, a total phosphorus concentration of 0.67 mg g^-1, and an organic matter concentration of 39.8 mg g^-1. Three pots with plants were hung on bamboo racks (6 m in length, 2 m in width, and 3 m in height) at three water layers: 0.4, 0.8, and 1.2 m, and a total of nine pots were placed in each pond. Macrophytes with lower C: N ratios (N-enriched waters) may be more readily consumed by herbivorous or omnivorous fish (Dorenbosch and Bakker, 2011). Before the experiment published in Yu et al. (2015), we performed an experiment with the same treatments but without excluding fish. All the macrophytes had been removed by the fish at the beginning of the experiment. Therefore, to ensure sufficient exposure time of macrophytes to high ammonium stress, nets with a mesh size of 2 cm × 2 cm were fixed around the bamboo racks to prevent fish from entering. The experiment lasted around two and a half month, from 1 April to 12 June 2013.

All plants were harvested at the end of the experiment and washed with tap water. The number of shoots (N_Shoot) was counted and shoot height (H_Shoot) was measured. To prevent the accumulation of NH₄ in plant organs, macrophytes usually incorporate it into some nitrogenous compounds [mainly free amino acids (FAA)] and/or transport it out of the organs, which processes consume carbohydrate and energy. About 100 mg of fresh shoots were ground in 5 mL 10% acetic acid solution to measure the contents of FAA. The above solution was centrifuged at 10,000 g for 15 min. The supernatant was used to examine FAA using the ninhydrin colorimetric method with leucine as standard (Li et al., 2004).

The dry mass of shoots (DM_Shoot) was measured with an electronic balance (0.01 g, BL-2200H, Shimadzu Corporation, Japan) after drying at 80°C (Jinghong, DHG-9071A, Shanghai) for 48 h to constant mass. Relative growth rates (RGRs, day^-1) of the macrophytes were calculated using the formula:

where T is the experimental period, W_t is the shoot number/shoot height/shoot biomass at the end of the experiment, and W₀ is the initial shoot number/shoot height/shoot biomass of the macrophytes.

There was no water flow through the ponds during the experiments. Environmental parameters were measured every 20 days. Dissolved oxygen and water temperature at 0.4, 0.8, and 1.2 m from the water surface were measured in situ with a YSI ProPlus (Yellow Spring Inc., United States). The water temperature difference between 0.4 m (upper layer) and 1.2 m (lower layer) was less than 1°C. Light intensity was measured by luminometer (KONICA MINOLTA, T-10, China) at the air-water interface (just above the water surface) and 0.4, 0.8, and 1.2 m below the water surface. Water samples for chemical analysis were collected at 5 randomly chosen locations within each pond by integrating the water column with a tube sampler (height 1.5 m and diameter 10 cm). Chlorophyll a of phytoplankton (Chla_Phyt) was extracted with 90% acetone (at 4°C for 24 h) after filtration through GF/C filters (Whatman, GE Healthcare UK Limited, Buckinghamshire, United Kingdom), and absorbance was read at 665 and 750 nm, both before and after acidification with 10% HCl using a spectrophotometer. The Chla_Peri concentration was calculated following the equation in Zhang and Huang (1991):

where C is the concentration of Chla_Peri in mg m^-2; Eb is D-value of absorbance at 665 and 750 nm before acidification with HCl; Ea is D-value of absorbance at 665 and 750 nm after acidification with HCl; R = 1.7; K = 11.24; V_e is total volume of extract in mL; and S is surface area of the slides in m².

Total nitrogen and total phosphorus (TP) were determined spectrophotometrically after digestion with K₂S₂O₈ solution (Huang et al., 1999). Total ammonium (NH₄) was determined with Nessler’s reagent colorimetric method (ISO 5664, 1984; ISO 7150-1, 1984). NH₃ was calculated following the equation in Emerson et al. (1975) and Wood (1993):

where pKa = 0.09018 + 2729.92/(273.2 + Temp), NH₄ is the measured concentration of total ammonium (including both NH₃ and NH₄⁺), pH is the measured pH of the solution, and Temp is water temperature in °C. In Yu et al. (2017) and this study, NH₃ and NH₄⁺ were expressed as total ammonium. Highly significant positive correlations were found between NH₃ and NH₄⁺ in Yu et al. (2018) (r = 0.90, n = 9, p < 0.001) and in this study (r = 0.98, n = 9, p < 0.001), reflecting that the high concentrations of NH₃ were accompanied by high NH₄⁺ concentrations. Consequently, the concentration of total ammonium (NH₄) in the two experiments serves as a combined indicator of toxicity, whether due to NH_3, NH₄⁺, or both. Therefore, measured NH₄ was used to express toxicity in both Yu et al. (2017) and this study.

In order to study the possible effects of periphyton on plants, artificial substrates were used to monitor the growth of periphyton. Three glass slides were embedded into a box with an open top and bottom, which was hung within the canopy of growing plants. Every 15 days, the glass slides were gently removed from the box for laboratory measurements after which a new set of three glass slides was embedded in the macrophytes. The periphyton growing on the glass slides was gently removed by a soft brush and flushed with 25 mL distilled water followed by filtered water, after which the concentration of periphyton chlorophyll a (Chla_Peri) was determined via acetone extraction.

Median values were used for analyses to avoid deviation caused by extreme values of TN, TP, Chla_Phyt, etc. Pearson’s correlations were used to test for relationships between macrophyte variables and environmental variables. Variables that did not follow normal distributions (Shapiro-Wilk test, p < 0.05) were log₁₀ (x)-transformed and these included NH₄, Chla_Phyt, and Chla_Peri. STATISTICA 8.0 and Microsoft Excel 2013 were used for all data analyses. Plants in pond N2b were grazed by invading fish as the net fell down. We therefore only analyzed the data from the other nine treatments.

Both realized TN and ammonium (NH₄) formed a significant treatment gradient, with medians ranging from 1.0 to 34.0 mg L^-1 (Figure 1A) and from 0.2 to 25.6 mg L^-1 (Figure 1B), respectively. The total amounts of added fertilizer are shown in Figure 1C. Chla_Phyt also increased significantly, with medians ranging between 10.5 and 162.4 μg L^-1. Chla_Peri declined from the upper (3.3–29.3 mg m^-2) over the middle (1.8–17.0 mg m^-2) to the lower (1.0–9.6 mg m^-2) layers in most treatments, except N0.5a and N100b (Chla_Peri was a little bit higher in the middle layer than the upper layer in the two treatments). Pearson’s correlations among the representative variables are given in Table 1. A highly significant positive relationship was found between TN and NH₄ (p < 0.01). There was no relationship between TN and Chla_Phyt (Figure 2A). A significant negative correlation was found between TN and Chla_Peri in the upper layer (p < 0.05) (Table 1 and Figure 2B) and between Chla_Peri and Chla_Phyt in the middle (p < 0.01) and lower layers (p < 0.05) (Table 1 and Figure 2C). A significant positive correlation was found between TP and Chla_Phyt (p < 0.05) (Figure 2D), and a negative correlation between Chla_Peri and TP, in the middle and lower layer (p < 0.05) (Figure 2E). No significant correlation was found between TN and TP (Figure 2F).

Growth and physiological variables of plants are shown in Table 2. No significant correlations were found between NH₄ and the growth variables (Table 3). The scatterplots did not show any clear changes in growth variables with increasing NH₄ concentrations (Figures 3A–C). FAA showed a significant increasing trend with NH₄ when pooling data from all depths (p < 0.001) (Figure 3D). When regressing FAA with growth variables of plants, N_Shoot demonstrated a significant positive relationship with increasing FAA in the middle and lower layers (p < 0.05). When pooling the data from all three layers, similar results were obtained (Figure 3E). H_Shoot in the middle layer exhibited a significant relationship with FAA (p < 0.05) (Figure 3F), while DM_Shoot was not related to increasing FAA (Figure 3G).

Chla_Phyt showed significant negative correlations with N_Shoot and H_Shoot in the lower layer and with H_Shoot in the upper and middle layers (p < 0.05) (Table 3). The scatterplots also showed generally declining trends of growth variables with increasing Chla_Phyt, significantly so for N_Shoot in the lower layer, for H_Shoot in all three layers, and for all growth indices when pooling the data from all three layers (Figures 4A–C).

Chla_Peri showed significant positive correlations with N_Shoot and H_Shoot in the middle and lower layers, and with DM_Shoot in the middle layer (p < 0.05) (Table 3). The scatterplots also demonstrated clear increasing trends of growth variables with increasing Chla_Peri, significantly so for N_Shoot and H_Shoot in the middle and lower layers, for DM_Shoot in the middle layer, and for all growth indices when pooling the data from all three layers (Figures 4D–F).

Although the precipitation varied in the experiments with P. crispus and V. natans, we maintained a fixed water level by pumping water out of the ponds after rainfall events. The average Secchi depth (SD) was similar in the two experiments, indicating that the two experiments apart from the temperature (see Discussion) were comparable despite the fact that they were conducted in two different years.

The relationships between NH₄ and growth variables for P. crispus and V. natans in the lower layer (1.2 m), a natural growth depth in the field, are compared in Figure 5. We found that 67% of the relative growth rates of shoot number (RGR_NS) was higher for V. natans than for P. crispus, while 78% of the relative growth rates of leaf length (RGR_Length) and 56% of dry mass (RGR_DM) of P. crispus were systematically higher than those of V. natans (Figures 5A–C). The RGR_NS of P. crispus and of V. natans demonstrated no significant relationships with increasing NH₄ (Figure 5A). RGR_Length and RGR_DM of V. natans both displayed a significant declining trend with NH₄, while no such trend was found for P. crispus (Figures 5B,C).