Isoetid mediated radial oxygen loss prevents iron reduction and the related mobilisation of ammonium and methane: an experimental approach

Isoetids are slow growing ecosystem-engineers that maintain an oligotrophic environment through high radial oxygen losses (ROL) from their roots. This ROL could potentially reduce methane (CH4) emission by maintaining sediment redox conditions unfavourable for methanogenesis. Isoetids have declined in many European countries as an effect of eutrophication and inorganic carbon enrichment. To assess the effect of sudden disappearance of isoetids on biogeochemical carbon and nutrient cycling in organic rich sediments, we set up a greenhouse experiment. Intact vegetated sediment patches with Lobelia dortmanna and Littorella uniflora were collected from ten boreal Norwegian softwater lakes. From half of the patches, isoetids were manually removed. We determined the effect of removal on sediment redox potential, pore water chemistry and CH4 emission. In addition, germination success of L. dortmanna seeds was tested on the sediments where isoetids were removed. The absence of ROL after isoetid removal led to a drop in sediment redox potential, resulting in anaerobic breakdown of organic matter using alternative electron acceptors, especially iron(hydr)oxides. Consequently, increased porewater concentrations of ammonium, iron, inorganic carbon and CH4 were observed in sediments where isoetids had been removed compared to the isoetid vegetated sediments. Moreover, over 20 times higher emission of CH4 was observed from the sediments where isoetids had been removed. On the highly reduced sediments, the germination of L. dortmanna was hampered. When isoetids are unable to recolonise after disturbance events, it is expected that the absence of ROL eventually results in a long-term increase in lake productivity and CH4 emission.


Introduction
Isoetid plant species, such as Littorella uniflora and Lobelia dortmanna, dominate softwater lakes in boreal and temperate regions and at higher elevation in the subtropics. Isoetid species are small, slow Abstract Isoetids are slow growing ecosystem-engineers that maintain an oligotrophic environment through high radial oxygen losses (ROL) from their roots. This ROL could potentially reduce methane (CH 4 ) emission by maintaining sediment redox conditions unfavourable for methanogenesis. Isoetids have declined in many European countries as an effect of eutrophication and inorganic carbon enrichment. To assess the effect of sudden disappearance of isoetids on biogeochemical carbon and nutrient cycling in organic rich sediments, we set up a greenhouse experiment. Intact vegetated sediment patches with Lobelia dortmanna and Littorella uniflora were collected from ten boreal Norwegian softwater lakes. From half of the patches, isoetids were manually removed. We determined the effect of removal on growing, evergreen water plants with an extensively developed root system (Hutchinson 1975;Roelofs et al. 1984;den Hartog and van der Velde 1988;Nielsen and Sand-Jensen 1997). During the past century, isoetid vegetations have declined considerably in many European countries due to eutrophication and inorganic carbon enrichment (Roelofs 1983;Arts 2002;Lucassen et al. 2016;Klimaszyk et al. 2020). Higher nutrient and inorganic carbon availability increases competition by favouring faster growing rooted macrophytes (Grahn 1977;Roelofs 1983;Lucassen et al. 2009) and alters the morphology of isoetids. It leads to an increased shoot weight and decreased root to shoot ratio, which increases the risk of plants becoming uprooted (Spierenburg et al. 2013;Lucassen et al. 2016). Mass uprooting events have already been observed in softwater lakes, for example in the Netherlands and Norway due to storm events (Spierenburg et al. 2013) and bottomfreezing ( Fig. 1), respectively. Isoetids are considered ecosystem-engineers and their sudden disappearance may strongly affect sediment processes in these softwater lakes.
The isoetid environment primarily consists of softwater lakes situated on poorly buffered siliceous bedrock or on non-calcareous sandy soils (Garcia et al. 1994;Murphy 2002). These lakes are typically fed by rainwater and shallow groundwater resulting in nutrient-poor, weakly buffered conditions with very low concentrations of dissolved inorganic carbon (Smolders et al. 2002a). In these carbon limited surface waters, carbon dioxide (CO 2 ) levels in the sediments are usually up to a 100-fold higher than in the water layer (Smolders et al. 2002a). Therefore, in contrast to most other aquatic macrophytes, isoetid species have developed the ability to take up CO 2 from the sediment for photosynthetic use (Boston et al. 1987;Nielsen et al. 1991;Pedersen and Sand-Jensen 1992;Madsen et al. 2002). Specific adaptations of the plants that enable CO 2 uptake from the sediments include a thick leaf cuticle, often without stomata, a well-developed root system (high root to shoot ratio) and the presence of a continuous system of gasfilled interconnected lacunae that run throughout the entire plant. These adaptations enable the effective diffusion of CO 2 from the sediment into the plants where the loss of CO 2 to the water layer is prevented by the leaf cuticle Smolders et al. 2002a). These Massive uprooting of the isoetids Lobelia dortmanna and Littorella uniflora in a pristine shallow Norwegian softwater lake during winter 2009-2010. This was probably caused by bottomfreezing. Extreme cold and dry periods stimulate the development of large areas with an ice cover filling in the complete water layer at the shores of the lakes, increasing the risk of isoetid uprooting morphological adaptations that allow isoetids to take up CO 2 from the sediment also cause a high radial oxygen loss (ROL) from the roots. Isoetids plants are therefore poorly adapted to grow on (highly) reducing sediments as the strong oxygen sink formed by these sediments would lead to a rapid extraction of oxygen from the permeable isoetid roots, thereby endangering the oxygen supply to the apical parts of the roots. Isoetid ROL stimulates aerobic mineralisation, thereby not only providing additional CO 2 for plant growth, but also promoting nitrogen (N) losses and immobilisation of phosphorus (P) (Christensen 1997;Risgaard-Petersen and Jensen 1997;Smolders et al. 2002a). Additionally, soil oxygenation can mitigate greenhouse gas (GHG) emissions by reducing methanogenesis and promoting CH 4 oxidation (Fritz et al. 2011;Ribaudo et al. 2011).
Freshwater lakes can be major sources of GHG emissions via microbially mediated anaerobic decomposition of organic matter (Bastviken et al. 2011). Northern freshwater lakes and ponds (> 50°N) are a considerable source of atmospheric CH 4 because of their large areal extent (Wik et al. 2016). The majority of this CH 4 emission originates from the littoral zone. In a study with three Finnish boreal lakes, Juutinen et al. (2003) found that the vegetated littoral zone of the lakes were major sources of CH 4 (66-77%) during the ice-free period. This can be attributed to submerged vegetation which supplies organic matter to the sediment as a substrate for methanogens (Chen et al. 2009;Xiao et al. 2017;Liu et al. 2019). Isoetids mainly occur in the littoral zones of oligotrophic softwater lakes (Smolders et al. 2002a) and therefore they may play a major role in preventing CH 4 emission from softwater lakes by oxidizing the sediment. In a field study, Ribaudo et al. (2017) indeed found limited CH 4 emission from isoetid vegetated sediments compared to bare sediments in two oligo-mesotrophic shallow lakes with mineral sediments and a low organic matter content (< 3%).
Although isoetids are most common on organically poor, mineral sediments they also occur on more organic soils (Woodhead 1951). In Norwegian softwater lakes, for instance, isoetids grow on sediments with organic matter contents up to 20-40% (Roelofs et al. 1994;Lucassen et al. 2016). These soils consist of relatively inert organic matter with a very low availability of nutrients and are a weak sink for oxygen. On these organic sediments, isoetids have the same growth form as on mineral sediments with small shoots and a relatively extensive root system. However, the risk of isoetid uprooting (due to disturbance events) is higher on more organic sediments due to the lower cohesive strength of these species in loose organic sediments (Spierenburg et al. 2013). Reduction of the organic rich sediments after isoetid disappearance and the absence of ROL may have a greater effect on CH 4 emission compared to sandy sediments, especially since CH 4 production and emission are stimulated by higher organic carbon availability (e.g. Ribaudo et al. 2011).
To assess the role of isoetids in the biogeochemical carbon and nutrient cycling, we collected intact vegetated sediment patches with L. dortmanna and L. uniflora from ten shallow boreal Norwegian softwater lakes with high sediment organic matter contents (10-66%). In a greenhouse experiment, we studied the effect of sudden isoetid disappearance by manually removing the isoetid vegetation from half of the patches. We determined the effect of this removal on sediment N, P, inorganic carbon and CH 4 availability and CH 4 emission. Moreover, to determine whether softwater lakes that have lost their isoetid vegetation could theoretically recover, we tested germination success of L. dortmanna seeds on the bare sediments where isoetids were manually removed. We hypothesised that isoetid removal promotes anaerobic breakdown of reactive organic matter due to the absence of ROL by isoetid roots, resulting in increased (inorganic) carbon and nutrient concentrations. Moreover, we hypothesised that the absence of ROL after isoetid removal creates favorable conditions for methanogenesis while reducing the CH 4 oxidation potential, thereby increasing production and emission of CH 4 from these sediments.

Isoetid removal experiment
In spring 2010, intact vegetated sediment patches with isoetids (L. dortmanna: 119 ± 65 g DW/m 2 ; L. uniflora: 12 ± 31 g DW/m 2 ) were collected in ten shallow, softwater lakes located in the southwestern part of the Norwegian province Rogaland (Suppl. Figure 1). The lakes were all located along the roads Nodlandsveien, Heggdalsveien and Vind-Birkedalsveien (except for the Fagervatnet lake). Most of this area is made up of silicious rocks with a thin layer of peat. All lakes were small (50-400 m 2 ) and contained a permanent, non-stratified water column. Intact vegetated sediment patches were collected from the littoral zone (max. 5 m from the shore). Here, the water level was approximately 20-60 cm. The patches were collected in duplicate (in total 20 patches) with a spade and cut to fit into dark plastic containers (length:widht:height = 35:25:8 cm ). Two additional samples of the sediment top layer (0-20 cm) were collected, close to where the intact vegetated sediments were collected, to determine sediment composition and stored in a dark and cold place (4 °C) until analysis ( Table 1). The 20 containers with intact vegetated sediment patches were transported in the dark to the Netherlands and placed in the greenhouse facilities of Radboud University Nijmegen with 16/8 h light/dark conditions and a temperature of 16-20 °C. Here, all plants were carefully removed from half of the containers (bare sediments; n = 10) while the isoetid vegetation in the remaining containers was left intact (vegetated sediments; n = 10). In all containers, two rhizon sediment porewater samplers (length 10 cm; Eijkelkamp Agrisearch, Giesbeek, The Netherlands) were installed diagonally in the sediment. The containers were placed into larger transparent plastic basins (length:weight:height = 40: 30:30 cm) and inundated by adding artificial rainwater. The artificial rainwater solution contained 5 mg/l sea salt ('Marine mix + Bio-elements', Wiegandt GmbH, FRG), 30 μmol/l KCl, 10 μmol/l CaCl 2 , 10 μmol/l Fe-EDTA, 0.7 μmol/l ZnSO 4 , 0.8 μmol/l MnCl 2 , 0.2 μmol/l CuSO 4 , 0.8 μmol/l H 3 BO 3 and 0.008 μmol/l (NH 4 ) 6 Mo 7 O 24 . During April 2010 (T = 0 weeks) and December 2010 (T = 33 weeks), sediment porewater was collected monthly (9 times) using the rhizons installed in the 20 containers. This was done by connecting the rhizons to vacuumed syringes. Together with sediment porewater collection, redox potential of the sediment top layer was also measured per container (for details, see "Chemical analyses"). CH 4 concentrations in the porewater were also measured monthly in pre-vacuumed glass bottles which were allowed to fill for 10% with porewater. Then, the vacuum was removed by filling the headspace with oxygen free nitrogen gas until normal pressure was reached. After allowing 4 h for equilibrating the headspace and water layer, CH 4 was measured in the headspace. Porewater composition showed that a new equilibrium had established three months after isoetid removal. We therefore only use sediment porewater composition from the last six months of the experiment (July-December 2010; 6 timepoints). These monthly porewater values were averaged and used to find correlations with redox sensitive parameters and were compared between vegetated and bare sediments. 1.1 1.9 0.3 3.7 28 1.3 1.6 0.7 9.6 1.0 4.9 ± 9 Tot-Zn (mmol/kg) 0.7 1.1 0.6 1.9 1.5 0.9 1.4 1.1 0.1 0.2 1.0 ± 1 CH 4 emission was quantified at the end of the isoetid removal experiment (December 2010) using upside-down funnels (radius = 5 cm; height = 4 cm) closed on top with a rubber stopper (Smolders et al. 2002b). Gas samples were collected at the start (T = 0) and at the end of the incubation period (T = 24; approx. 24 h) through the rubber stopper by using a syringe (1 ml) and a needle. The CH 4 emission was calculated from the difference in headspace concentration in the funnel between T = 0 and T = 24, which was multiplied by funnel volume and divided by funnel area to establish the flux in mg CH 4 -C/m 2 / day.

Germination experiment
At the end of the experiment, a plastic open ring (radius = 4 cm; height = 4 cm) was partly pushed into the top layer of the sediment of the containers from which the vegetation had been removed (n = 10), after which 40 L. dortmanna seeds were added to each ring. The percentage of seed germination was monitored by counting the number of germinated seeds seven times between January 2011 and the beginning of March 2011. Redox potential of the sediment top layer was measured within the rings in March 2011. At the end of the experiment, germination success was defined as the maximum germination percentage of the seeds per container. This value was plotted against the corresponding redox potential of the sediment top layer.

Chemical analyses
Sediment composition was determined from the samples that had been collected separately at each of the ten lakes. Organic matter content was determined as loss on ignition, by combusting a known amount of dried (homogenised) sediment for 4 h at 550 °C. Sediment density and moisture content were determined by drying a known volume of fresh sediment for 48 h at 60 °C. Sediment elemental composition was determined by digestion of 200 mg of homogenised, dried sediment with 4 mL of nitric acid (65%) and 1 mL of hydrogen peroxide (30%) using a microwave (Milestone microwave type mls 1200 mega), after which samples were analysed for total (tot) aluminium (Al), iron (Fe), P, calcium (Ca), potassium (K), sulfur (S), silicon (Si), magnesium (Mg), manganese (Mn), and zinc (Zn) content by Inductively Coupled Plasma (ICP) spectroscopy (IRIS-OES model Intrepid II XDL; Thermo Fisher Scientific,Waltham, MA, U.S.A). Plant-available P (Olsen-P) was determined by soil extraction after shaking 3 g of dried sediment with 60 mL of sodium carbonate (0.5 mol/l) for 30 min. Extracts were analysed for tot-P content using ICP.
Redox potential was measured at 5 cm depth in the sediment with a multimeter (p901, Consort, Belgium), a platinum electrode and an Ag/AgCl reference electrode (Metrohm, Switserland). The measured electrical potentials were converted to redox potentials relative to the standard hydrogen potential (Eh) by adding the reference (210 mV) and correcting for temperature and pH of the porewater. The pH of the collected porewater was determined using a titration workstation (TitraLab 840, Radiometer analytical SAS, Villeurbanne, France) with a double Ag/AgCl pH electrode (Orion 9156BNWP, Thermo Scientific, USA). The concentration of total inorganic carbon (TIC) in the porewater was measured on an infrared carbon analyzer (Advance Optima, ABB, Cary, NC, U.S.A.). Based on the pH and TIC concentrations, the CO 2 and bicarbonate (HCO 3 − ) concentrations were calculated according to Stumm et al. (1996). The tot-P, S, Fe content of the porewater samples from the bare and vegetated sediments were determined in acidified samples (1% nitric acid) using ICP. The porewater ammonium (NH 4 + ) and nitrate (NO 3 − ) content were analyzed by an Auto Analyser system (model III, Bran and Luebbe, Nordstedt, Germany) using salicylate and hydrazine sulphate, respectively. CH 4 gas samples were analysed with ethane as an internal standard on a Pye Unicam gas chromatograph Porapak Q (80/100 mesh) column (Water Chromatograpy, Etten-Leur, The Netherlands).
Chemical analyses were conducted by the General Instrumentation (GI) of the Faculty of Science of the Radboud University. All values were above the detection limit.

Statistical analyses
The differences in mean pore water chemistry (mean of 6 monthly measurements) and CH 4 emission (measured at the end of the experiment) between vegetated (n = 10) and bare sediments (n = 10) was tested using linear mixed effect models with isoetid removal treatment (vegetated/bare sediments) as fixed factor and lake as random effect (to avoid violation of independence). Regressions between sediment porewater chemistry and redox sensitive parameters (i.e. redox potential; porewater Fe 2+ /TIC concentrations) and between CH 4 emission and porewater CH 4 concentration were tested using linear models separately for the vegetated and bare sediments. Linear models assumptions were satisfied for normality (Kolmogorov-Smirnov test) and heteroscedasticity (Levene's test). All data were analysed using Microsoft Excel 2016 and R software version 3.4.3 (R Core Team 2017). For R, the following packages were used: "nlme" (Pinheiro et al. 2021, version 3.1-155) and "ggplot2" (Wickham 2016, version 3.3.5). Means with standard deviation (SD) are given throughout the text. Statistical significance was accepted at p ≤ 0.05.

Germination success L. dortmanna
Germination of L. dortmanna varied between 0 and 70% of the added seeds and only occurred on sediments with a redox potential above -4 mV (Fig. 5).

Discussion
In our experiment, we revealed that the manual removal of isoetids led to a drop in redox potential and to higher inorganic carbon, inorganic N and CH 4 concentrations in the sediment porewater (Fig. 6). CH 4 emission also strongly increased with the removal of isoetids. In strongly reduced sediments, the germination of isoetids was hampered.

The effects of isoetid removal on inorganic carbon and nutrient availability
Due to the high ROL along their entire root system, isoetids have a strong influence on nutrient and carbon cycling in softwater lakes (Wium-Andersen and Andersen 1973;Smolders and Roelofs 1996;Pedersen et al. 2011). Oxygenating the sediment elevates the redox potential, even in more organic systems, thereby stimulating N losses Fig. 3 Correlations between porewater (pw) Fe and inorganic carbon (a, b), NH 4 + (c) concentrations and between tot-Fe content (sediment) and porewater Fe concentration (d) of the isoetid vegetated sediments (white dots) and sediments where isoetids had been removed (black dots). Porewater values rep-resent the mean (± SE) of 6 measurements conducted monthly between July and December 2010 (3-9 months after isoetid removal). Note that tot-Fe content was measured only once at the start of the experiment (April 2010). Lines represent linear regressions and the P binding capacity of the sediment (Smolders et al. 2002a;Lucassen et al. 2009).
Our results show that the absence of ROL following manual removal of isoetids leads to anaerobic sediment conditions and a drop in redox potential. Without ROL, labile organic matter is oxidised through alternative terminal electron acceptors, i.e. NO 3 − , Mn(IV)oxides, Fe(III)(hydr)oxides or SO 4 2− (Knorr and Blodau 2009). During this anaerobic degradation of labile organic matter, CO 2 is produced which dissolves in the porewater and partly reacts to form carbonic acid (H 2 CO 3 ). Reduction processes with alternative electron acceptors (e.g. Fe(III) (hydr)oxide) consume protons, which are split off from H 2 CO 3 , thereby resulting in a raised pH and a net production of HCO 3 − (Schindler 1988;Lucassen et al. 2009). This process is known as internal alkalinisation (Roelofs 1991). In reducing sediments of limed Norwegian softwater lakes, mainly Fe(III)(hydr)oxides acted as the dominant electron Correlations between porewater (pw) CH 4 concentration and redox potential (a), TIC (b) and between CH 4 emission and the concentration of porewater CH 4 (c) of the isoetid vegetated sediments (white dots) and sediments were isoetids had been removed (black dots). Porewater values and redox potential represent the mean (± SE) of 6 measurements conducted monthly between July and December 2010 (3-9 months after isoetid removal). Note that CH 4 emission was only measured once at the end of the experiment (December 2010). Lines represent linear regressions acceptor, resulting in Fe mobilisation (Roelofs et al. 1994;Lucassen et al. 2009Lucassen et al. , 2016. The reduction of Fe(III)(hydr)oxides under anaerobic conditions was also evident in our experiment, with significantly increased ferrous iron (Fe 2+ ) concentrations after isoetid removal. Furthermore, we found a strong negative correlation between Fe 2+ concentrations and redox potential in the bare sediments. When Fe is reduced, previously Fe-bound P can be mobilised in the sediment porewater (Tessenow and Baynes 1975;Smolders et al. 2006;Lucassen et al. 2009). We only found a small increase in P concentration after isoetid removal. This can be explained by the high tot-Fe content of the sediments, which can still bind substantial amounts of P.
Besides P mobilisation, iron-mediated anaerobic degradation of organic matter also results in the conversion of organic N to inorganic N in the form of NH 4 + (de Jong et al. 2020). Due to the lack of oxygen after removal of isoetids, NH 4 + accumulates in the porewater since it is not oxidised to NO 3 − . As a result, coupled nitrification-denitrification is limited in the absence of ROL (Risgaard-Petersen and Jensen 1997;Smolders et al. 2002a). In softwater lakes with relatively organically rich sediments, in-lake alkalinity generation may stimulate organic matter breakdown, contributing to even higher NH 4 + concentrations in the sediment porewater ). These processes explain the strong increase in NH 4 + concentration in the sediments where isoetids had been removed. The positive correlation between porewater NH 4 + and Fe 2+ concentrations in the bare sediments reveals that ferric iron (Fe 3+ ) is the main electron acceptor under anaerobic conditions. These results are in line with other studies showing raised carbon and nutrient availability and increased Fe 2+ concentrations after a drop in oxygen concentrations in the sediment, for example due to absence of ROL after massive uprooting of isoetids on organic soils   . 6 Schematic overview of the effects of isoetid vegetation (removal) on biogeochemical carbon (including CH 4 ) and nutrient cycling in organically rich sediments. Arrows with " + " indicate stimulation, arrows with "−" indicate suppression (Spierenburg et al. 2013) or after rewetting of organically-rich (peat) soils (de Jong et al. 2020).

The effects of isoetid removal on CH 4 production and emission
Manual removal of isoetids resulted in increases in porewater CH 4 concentrations. In addition, at the end of the experiment, we measured more than 20 times higher CH 4 emission from the bare sediments compared to the isoetid vegetated sediments.
With the removal of isoetids and the absence of ROL, favourable conditions arise for methanogenesis (Bodegom and Stams 1999;Lai 2009). In wetlands, CH 4 is mainly produced by methanogenic archaea following microbially mediated anaerobic decomposition of organic matter via the conversion of acetate (acetoclastic methanogenesis) or CO 2 and hydrogen (hydrogenotrophic methanogenesis) to CH 4 (Schink 1997;Bastviken et al. 2011). The produced CH 4 can be emitted to the atmosphere via plant-mediated transport, ebullition or diffusion, or it can be oxidised by methanotrophs in aerobic niches of the soil or in the water layer (Semrau et al. 2010;Torres et al. 2014;Dean et al. 2018). Microbial CH 4 oxidation is one of the most important factors controlling CH 4 emissions from lakes (Chistoserdova 2015), by oxidising (part of) the produced CH 4 in the sediment before it is emitted to the atmosphere. This may play a role in the fact that a relatively weak correlation between porewater CH 4 availability and CH 4 emission (measurement at the end of the experiment) was observed in the bare sediments. Methanogenesis generally takes place after depletion of other alternative electron acceptors when the production of CH 4 becomes an energetically favourable respiration pathway (Teh et al. 2007;Yamada et al. 2014). Nevertheless, substantial CH 4 porewater availability and emission were measured from the iron-rich sediments used in our experiment, which implies that methanogenesis was not inhibited by the presence of Fe oxides. Similar results were found by de Jong et al. (2020) for rewetted peat soils. This can be attributed to the fact that methanogenesis can still take place in microenvironments in the upper sediment where substantial amounts of CH 4 can build up before alternative electron acceptors are entirely consumed (Knorr et al. 2008;Knorr and Blodau 2009) and/or that Fe reducing bacteria and methanogenic archaea do not entirely compete for the same organic substrates.
Anaerobic respiration and related CH 4 production is remarkably reduced with the continuous and thorough soil oxygenation, for example by isoetids (Pinardi et al. 2009;Ribaudo et al. 2011Ribaudo et al. , 2017. Oxygenation of the sediment through ROL stimulates methanotrophic activity, leading to oxidation of CH 4 produced in anaerobic soil niches (Popp et al. 2000) and to reoxidation of reduced alternative electron acceptors, thereby suppressing CH 4 production (Laanbroek 2010). A similar result was observed in pristine bogs in Patagonia with a high density of cushion bog plants, which thoroughly oxidised the soil (Fritz et al. 2011). We measured higher CH 4 concentrations after isoetid removal compared to bare sandy soils in the study by Ribaudo et al. (2017). This might be attributed to the generally higher organic matter content of the sediments in our experiment. CH 4 production and emission are stimulated by higher availability of reactive organic carbon, resulting in higher methanogenic activity (Ribaudo et al. 2011;Grasset et al. 2018).

Implications
Our microcosm study confirms that isoetids are effective in maintaining oligotrophic conditions and mitigating CH 4 emission, even in softwater lake sediments with a higher organic matter content. When isoetids disappear through disturbances following eutrophication or inorganic carbon enrichment, sediments become reductive (Fig. 6). This hampers recolonisation of isoetids due to low germination success (Arts et al. 1990;Bellemakers et al. 1996). We also found that L. dortmanna failed to germinate on sediments with a redox potential < − 4 mV. The low germination success of L. dortmanna seeds can be caused by oxygen stress on more reduced sediments. For L. dortmanna, it is known that anoxia can induce a secondary dormancy preventing seed germination (Farmer and Spence 1987). Other isoetid species, such as L. uniflora, require emerged conditions for germination (Arts et al. 1990). Even if isoetids would germinate on reductive sediments, long periods of anoxia can result in isoetid mortality (Sand-Jensen et al. 2005;Lucassen et al. 2016). If isoetids are unable to re-establish on reductive sediments, more productive species may take over permanently, such as Juncus bulbosus and Sparganium angstifolium (Smolders et al. 2002a;Spierenburg et al. 2009Spierenburg et al. , 2013Lucassen et al. 1999Lucassen et al. , 2011Lucassen et al. , 2016. When fastergrowing species establish permanently, these systems can switch from oligotrophic softwater lakes to more mesotrophic lakes. CH 4 emission from lakes is strongly controlled by the nutrient status of the lake (Rasilo et al. 2015;DelSontro et al. 2018;Santoso et al. 2021). Eutrophication stimulates plant production, which leads to increased organic matter input into the sediment, higher oxygen consumption and eventually higher CH 4 emission (Davidson et al. 2015;Xiao et al. 2017;Beaulieu et al. 2019;Vachon et al. 2020). Compared to our experiment, Kankaala et al. (2003) found up to 60 times higher CH 4 emission from a meso-eutrophic boreal lake with nonisoetid vegetation. Besides higher lake productivity, this can also be caused by increased plant-mediated transport of CH 4 to the atmosphere bypassing methanotrophic zones of the soil (Carmichael et al. 2014). Yang et al. (2015) also found that CH 4 emission was higher from mesotrophic boreal lakes compared to oligotrophic boreal lakes, with similar CH 4 emission rates from oligotrophic lakes compared to the sediments from which isoetids had been removed in our study.
Considering that isoetids are declining in many regions due to environmental changes and that recolonisation success is often low, longterm absence of isoetids may lead to increased productivity of these originally oligotrophic softwater lakes causing these systems to transform into net CH 4 sources.
Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

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

Declarations
Conflict of interest All authors declare that they have no conflicts of interest.