Land cover and nutrient enrichment regulates low-molecular weight dissolved organic matter turnover in freshwater ecosystems

Dissolved organic matter (DOM) is a complex mixture of carbon-containing compounds. The low-molecular weight (LMW) fraction constitutes thousands of different compounds and represents a substantial proportion of DOM in aquatic ecosystems. The turnover rates of this LMW DOM can be extremely high. Due to the chal-lenges of measuring this pool at a molecular scale, comparatively little is known of the fate of LMW DOM compounds in lotic systems. This study addresses this knowledge gap, investigating the microbial processing of LMW DOM across 45 sites representing a range of physicochemical gradients and dominant land covers in the United Kingdom. Radioisotope tracers representing LMW dissolved organic carbon (DOC) (glucose), dissolved organic nitrogen (DON) (amino acid mixture), dissolved organic phosphorus (DOP) (glucose-6-phosphate), and soluble reactive phosphorus (SRP, measured as orthophosphate) were used to measure the microbial uptake of different DOM compounds in river waters. The amount of DOM biodegradation varied between different components (DON ≥ DOC > DOP), with

Dissolved organic matter (DOM) is a complex mixture of chemicals, traditionally defined as organic carbon (C)containing compounds that can pass through a 0.45 μm filter (Thurman 1985;Akkanen et al. 2012). DOM constitutes a key form in which terrestrially derived C is transported from headwaters, through the catchment hydrological network and into the marine environment. It has been estimated that~33% of terrestrial C is exported to the ocean in this way (Stutter et al. 2018). In addition to C, DOM also contains nitrogen (N) and phosphorus (P), which are together considered to be the three major nutrients required by freshwater organisms (Tipping et al. 2016). DOM therefore represents a key source of nutrients for microorganisms and plants along the aquatic gradient (Kirchman 2003;Cuss and Guéguen 2015). Further, recent research has shown that DOM quality and quantity can change from source to sea suggesting that components of the DOM pool are abiotically and biotically transformed during transit (Battin et al. 2008;Massicotte and Frenette 2011;Ejarque et al. 2017).
DOM is a highly complex mixture composed of tens of thousands of individual compounds differing in size, charge, and solubility. Broadly, DOM can be divided into two size categories, namely those compounds that are of high molecular weight (HMW; > 1000 Da) and those that are of low molecular weight (LMW; < 1000 Da) (Cui and Choo 2013). High-molecular weight DOM, typically lignin and lignin-derived breakdown products from woody debris, is considered to be relatively recalcitrant and not readily degraded by the microbial biomass (Zhang et al. 2017). Therefore, it is often present at high concentrations, particularly in peat-rich headwater catchments. It can, however, have an abiotic function, reducing light attenuation in the water column due to its chromophoric properties, shielding the microbial biomass, and extracellular compounds from UV degradation (Fellman et al. 2010) and is readily broken down through photodegradation to release lower MW compounds to the river ecosystem as the material moves downstream (Cory and Kling 2018;Bowen et al. 2019). In general, LMW DOM (e.g., sugars, organic acids) is present in lower amounts (ca. 20% of the total DOM pool) in rivers draining from natural and seminatural landscapes, rising in waters draining through urbanized and intensively farmed agricultural landscapes (Yates et al. 2019b) but is often more labile and rapidly metabolized by the microbial biomass (Dawson et al. 2001). This makes the measurement of this fraction both important, as the LMW DOM pool is an important component of instream C, N, and P processing and biotic uptake, as well as technically challenging (Lutz et al. 2011;Salcher et al. 2013;Parr et al. 2014). In natural and seminatural systems, individual LMW DOM compounds (e.g., amino acids) are present at very low concentrations in freshwaters (1-10 nmol L À1 ; Marie et al. 2015;Hor n ak et al. 2016). These concentrations are close to the influx-efflux equilibrium point for microbial transport systems (i.e., point of zero net flux). It is unclear, however, if this is a reflection of low rates of LMW DOM input or whether it is due to high rates of consumption, especially for a wide range of compound classes and freshwater types.
In contrast, as systems become nutrient enriched through anthropogenic activity LMW DOM compound concentrations increase, this will exceed the equilibrium point and stimulate higher rates of microbial consumption (Sirivedhin and Gray 2005). The aim of this study was therefore to: (1) compare the rate of microbial uptake of LMW forms of DOC (sugars), DON (amino acids), and DOP (sugar phosphates) at different times of the year in 45 individual rivers draining through landscapes with a range of physicochemical properties and land covers; (2) determine which physicochemical parameters best correlated with DOC, DON,and DOP uptake,and (3) compare the rates of DOP and inorganic P (SRP) use by the microbial community. The results of the study were then used to evaluate how DOM and SRP are processed across catchment-scale gradients of disturbance and nutrient enrichment.

Field site and sampling
Samples were collected from 45 rivers across five contrasting land covers (arable, grasslands, conifer forests, peatland, and mixed agricultural) in the Conwy catchment, North Wales, and the Nadder Catchment, a subcatchment of the well-researched Hampshire Avon catchment in Southern England (Jarvie et al. 2005;Pirani et al. 2016;Fig. 1). This was undertaken alongside regular monitoring of all sites between February 2015 and December 2016, detailed methods are give in Yates et al. (2019b). The sites selected encompass a wide range of chemical and physical gradients and contain a range of dominant land covers spanning from natural and seminatural ecosystems to moderately and intensively farmed improved agricultural grassland with settlements discharging septic tank and sewage effluent to the rivers (Emmett et al. 2016;Lloyd et al. 2019;Yates et al. 2019a).
The Conwy catchment (Fig. 1a) covers an area approximately 580 km 2 , draining a wide range of land-cover types, and is underlain by Silurian siltstones, mudstones and slate to the east and older, more resistant Ordovician rocks to the west. The lithology has a low permeability and soils are acidic. Land cover in this part of the catchment comprises the Migneint, one of the largest areas of upland blanket peat bog in Wales in its headwaters, together with acid grasslands, coniferous plantations, and broadleaf forests. The lower half of the catchment is underlain by Ordovician mudstones, siltstones, and sandstone with a lower relief, with acidic to neutral soils supporting lowland improved grassland used for moderately intensive agricultural production primarily sheep farming, before entering the Irish Sea (Cooper et al. 2014;Emmett et al. 2016;Brailsford et al. 2019a). The Nadder catchment covers a larger area (673 km 2 ; Fig. 1b). In its headwaters, the Nadder is underlain by Gault Clay and Upper Greensand, supporting intensive cattle production on its neutral to slightly alkaline clay soils, while the Wylye, its major tributary, is underlain by Cretaceous Chalk, is heavily influenced by groundwater recharge, and supports intensive arable production on calcareous brown earths (Yates and Johnes 2013;Yates et al. 2019b). This range of landscape characteristics and resulting land cover and use provided a gradient of sites ranging from C-rich, highly acidic soils to mineral calcareous soils, from high natural organic matter (NOM) to low NOM systems, and from low N and P to high N and P flux in either dissolved organic, particulate, or soluble inorganic form. Land cover similarly varied from woodland to heathland, and acid grassland to intensively fertilized and manured improved grassland, and mineral soils supporting arable production with no crop cover for much of the year. Each would generate a different DOM profile from its contributing sources to the rivers draining each landscape type (Yates et al. 2019b). Sites for this suite of experiments were selected to reflect these gradients. Details on site characteristics are given in the Supporting Information Table S1.
At each site, 1 L midstream samples were manually collected in acid-washed, high density polyethylene bottles. The pH, electrical conductivity (EC), and water temperature were measured at the time of collection. For the laboratory studies, the samples were kept cool (ca. 4 C) in the dark from the point of collection and during transportation to the laboratory where they were placed in cold storage. Chemical analyses were conducted, and experiments commenced within 24 h of sample collection (see Yates et al. 2019b for detailed methods).

Nutrient depletion experiment
To evaluate DOC, DON, DOP, and SRP depletion across the samples, individual samples were spiked with either: (1) 14 Clabeled glucose; (2) a mixture of 14 C-labeled free amino acids; (3) 14 C-labeled glucose-6-phosphate; or (4) 33 P-labeled PO 4 3À , respectively. For each treatment, three independent replicate 25 mL samples from each of the 45 sampling sites were added to sterile 50 mL polypropylene centrifuge tubes (Corning, New York, U.S.A.) and spiked with individual radioisotope (0.2 mL, 0.2 kBq mL À1 final activity). The radioisotopes were added at low concentrations (< 1 nmol L À1 ) which would be unlikely to increase the background pool of the target compound or change its pH (Brailsford et al. 2017). After sealing with sterile caps, samples were incubated on an orbital shaker (200 rev min À1 ) in the dark at 10 C for the duration of the experiment. After incubation for 2, 5, 24, 48, 72, 144, or 168 h, 1 mL subsamples were removed, centrifuged to remove microbial cells (20,817 Â g, 5 min), and 0.5 mL supernatant added to a plastic 7 mL scintillation vial (Meridian Biotechnologies, Tadworth., UK). The subsamples were then acidified with 0.1 mol L À1 HCl (50 μL), vortexed, left to stand for 3 h and then vortexed again to remove any dissolved 14 CO 2 present. The subsample was then mixed with Optiphase HiSafe 3 scintillation cocktail (4 mL; PerkinElmer, Waltham, Massachusetts, U.S.A.) and the 14 C or 33 P quantified on a Wallac 1404 liquid scintillation counter with automated quench correction (Wallac EG&G, Milton Keynes, UK).

Statistical analysis
The physicochemical parameters used in the analysis included DOC (mg C L À1 ), total nitrogen (TN) (mg N L À1 ), DON (mg N L À1 ), NO 3 -N (mg N L À1 ), NO 2 -N (mg N L À1 ), NH 4 -N (mg N L À1 ), total phosphorus (TP) (mg P L À1 ), DOP (mg P L À1 ), SRP (mg P L À1 , measured as PO 4 -P), pH, and EC (μS cm À1 ). All physicochemical parameters were tested for normality using the Shapiro-Wilk test. Nitrate and DOC were log 10 transformed to pass the normality test. The percentage depletion at 24 h was used for comparison as this captured the linear phase of the degradation curve for the majority of sites and radioisotope tracers used. A one-way ANOVA with Tukey pairwise comparisons was undertaken for the 24 h depletion data using Minitab v18.0 with p < 0.05 as the cutoff for statistical significance (Minitab, State College, Pennsylvania, U.S.A.). Canonical correspondence analysis (CCA) of physicochemical parameters for the 45 sites and mean radioisotope tracer depletion data was conducted using the envfit function in the Vegan package (R version 4.0.3; Oksanen et al. 2018; R Core Team 2019). Environmental data were scaled to unit variance when applying the CCA. To note, the results from the experiments repeated over the four seasons were averaged together as season was not found to be a major driver of the differences between sites during initial analyses and many sites did not exhibit distinct variation between seasons (Supplementary Document 1). The significance of the ordination was tested using a permutation test. The coordinates resulting from the CCA analysis were plotted using the ggplot2 package (Wickham 2016). Arrow/vector length and position provide information about the relationship between the environmental variables and the axes, with arrows parallel to axis orientation indicating a correlation and the length of the arrow inferring strength of that correlation. Permutational multivariate analysis of variance (PERMANOVA) was used to test the effect of land cover type on the dispersal of sites within the CCA. Prior to this, homogeneity of multivariate dispersions throughout the sample was confirmed using a beta dispersion test.

Physicochemical properties
Significant variation was apparent between the chemistry of the river waters across the 45 sites (Supporting Information - Table S1; Yates et al. 2019b). Comparison of the rate of substrate depletion over 24 h across all samples also showed differences in substrate depletion rate following the series: glucose ≥ amino acids ≥ glucose-6-phosphate > SRP (p < 0.001; Table 1; Supporting Information Table S2). In addition, there were major differences in substrate depletion between land cover for all three organic substrates (p < 0.001); however, no difference in SRP depletion was observed across the land covers (p = 0.613). Overall, rates of substrate depletion were greatest in catchments draining agricultural land (arable and grassland in both catchments) in comparison to sites draining peat moorland and coniferous forest in the Upper Conwy catchment. Notably, SRP depletion was only greater than DOP depletion rates in sites draining from the peat moorland and coniferous forest landscapes where P concentrations are naturally extremely low (Supplementary Document 1).
To identify the major factors associated with the depletion of isotope from solution, mean rates of initial DOC, DON, DOP, and SRP depletion were analyzed alongside physicochemical parameters of the water (Table 1; Fig. 2; Supporting Information Tables S1-S3). Two canonical axes (CCA1 and CCA2) explained 82.5% of the overall variation between the 45 sites (Table 2; Fig. 2). The overall model was statistically significant, with the CCA1 axis being a significant driver of differences between sites (permutation test, p = 0.001 for both).
Dominant land cover had a significant effect on sample groupings in CCA analysis; when plotted with coordinates resulting from the CCA ordination analysis, rivers draining peatland (HMW DOC rich, N/P depleted) formed a distinct cluster in the bottom right quadrant ( Fig. 2; Table 3). There was also clustering of some separation between catchments, with sites from the along the Wylye-Nadder catchment, clustering on the left hand side of the plot. In contrast, the majority of sites located within the River Conwy catchment clustered in the top right quadrant (Supporting Information - Table S3). However, improved grassland-influenced rivers (low DOC, N and P enriched) sites from the Conwy catchment rivers were more closely associated with Wylye-Nadder catchment rivers with a similar level of N and P enrichment, clustering together on the right of the x-axis (sites 2, 10, 20, 21; Supporting Information Table S3), suggesting that their behavior in terms of the utilization of DOM as a resource, is strongly controlled by the level of nutrient enrichment in the catchments.

Physicochemical and nutrient controls on processing in rivers
The results presented in this study indicate that both the physicochemical character of the river and its catchment, and its nutrient enrichment status strongly influence how LMW DOM components are biologically processed in freshwaters. Rates of DOM depletion observed in this study were highest in those sites with higher nutrient concentrations, draining from intensively farmed landscapes, and lowest for systems with high background DOC from peaty soils with low N and P concentrations, typically supporting blanket bog, acid grassland, and forestry. This finding confirms the findings of previous studies, including our own, which have shown that dominant land cover is an important determinand of DOM quality and stoichiometry in freshwaters, influencing both the C : N and DOC : DON/DOP ratios (Yates et al. 2019b) and that physicochemical factors control the rate of biodegradation of DOM in rivers (e.g., Guillemette et al. 2013;Sleighter et al. 2014;Catal an et al. 2017). Here, both the relative and absolute pool sizes of DOC, TN, and TP appeared to directly influence the separation of sites (Table 2; Fig. 2). Those sites with N and P pools enriched by either inorganic or organic or both forms of N and P show higher rates of depletion of dosed DOC, DOC, and DOP than sites with a lower background concentration of N and P. In this study, it is also clear that controlling factors are intertwined. Those sites in our sample of 45 locations which have lower background N and P concentrations also have a much higher proportion of HMW DOM relative to LMW DOM, based on optical measurements (Yates et al. 2019b). There was also correlation between land cover, underlying geology, soil acidity, and DOM character at each site, with mountain and moorland sites underlain by impermeable geology supporting acid grassland and peat bog with a higher acidity and higher proportion of HMW DOM along with very low background N and P concentrations. Thus, it is impossible in this experimental framework to entirely attribute the different rates of DOC, DON, and DOP depletion to any one controlling factor. As has been argued in many prior studies, multiple physicochemical and biological controls are exerted on the function of any one biotic group within the ecosystem, controlling the rate of DOM processing and uptake by riverine biota. What is clear from these findings, however, is that the nutrient enrichment status of a waterbody is an additional factor not typically considered in prior work, which also acts to control the rate of biodegradation Table 1. Percentage depletion of 14 C-labeled DOC, DON, and DOP and 33 P-labeled SRP from river water across five major land cover types after 24 h. Values represent percentage of tracer depleted within 24 h and are presented as means AE SEM. The average value across all the sites is also presented at the foot of the table. Significant differences between land use types are shown with lowercase superscript letters while differences between substrates are denoted by uppercase superscript letters (p < 0.05).

Land cover
Depletion 24   of labile DOM in river systems. We explore this evidence further below.
Degree of nutrient enrichment as a control of degradation In this study, the DOC, TN, and TP concentrations in the rivers sampled were correlated with the rate of microbial degradation of DOC, DON, DOP, and SRP (Table 2). Across the 45 sites, the rate of DOC depletion decreased with increasing background DOC concentration, although this may have been a function of the higher proportion of HMW : LMW DOM in the sites with the highest DOC concentrations. The highest observed DOC concentrations were in peatland influenced rivers, where TN and TP concentrations were among the lowest observed (Supporting Information Table S1; Emmett et al. 2016;Yates et al. 2019b). Of note, the proportion of the TN and TP that was in the form of DON and DOP (Supporting Information Table S1) in these sites was the highest, decreasing relative to inorganic N and P as TN and TP concentrations increased, in line with trends in N speciation reported across Europe by Durand et al. (2011). In these headwater sites, where C is abundant, but N and P are scarce, N/P colimitation is likely to be the main physicochemical factor controlling the microbial processing of LMW DOM. Here we observed rapid depletion of our DON and DOP substrates, and argue that this demonstrates the biotic uptake and metabolism of both as a nutrient resource in these nutrient-poor sites. For sites subject to N and P enrichment from human activities in their catchments, we observed an increase in DOC uptake by the microbial biomass, which can be ascribed to the removal of metabolic constraints associated with N/P limitation on microbial growth and therefore labile LMW DOC uptake (Carlson and Ducklow 1996;Creamer et al. 2014;Parr et al. 2015;Fovet et al. 2020). It has been argued that rivers transition from being N/P limited to N/C limited along a gradient from source to sea (Jarvie et al. 2018). However, this would not be a general truth in disturbed river catchments (farming, urbanization) where the nature and location of disturbance would be the key control on nutrient availability and stoichiometry at any particular location rather than any simple measure of distance downstream. Furthermore, it could be argued that increasing N and P enrichment of rivers could potentially increase autochthonous DOC, DON, and DOP synthesis, such that the associated enhancement of microbial growth and consequent increase in the rates of DOM degradation (as observed in the current study) can drive microbial communities toward C limitation (Stanley et al. 2011;Emmett et al. 2016). This has been observed in the current study, where labile DOC, DON, and DOP depletion in our bioassays increased as background N and P concentrations increased along our gradient of sites, confirming N/P enrichment (Table 2; Supporting Information Table S1). By contrast, SRP depletion from solution decreased with increasing nutrient concentrations; this was in agreement with our study of a nutrient-enriched riparian zone where the addition of SRP, did not lead to an enhancement in microbial activity (represented by 14 C uptake) due to a lack of P limitation, and thus a lack of demand for additional SRP (de Sosa et al. 2018).
Based on the regression analysis parameters, LMW DOM depletion followed the trend: DON ≥ DOC > DOP, even in low nutrient status waters (Table 2). A previous catchment-scale study of DOM metabolism found that DON (in the form of amino acids) degraded quickest in rivers draining through peatland compared to those influenced by other land covers, which is likely influenced by the N limitation of these ecosystems (Berggren and del Giorgio 2015). In this study and in our earlier work, we observed that the microbial degradation of amino acids in oligotrophic peatland rivers was slower than in mesotrophic grassland rivers, and that this amino acid depletion was quicker than that of glucose, organic acids, and phenolic compounds in both river waters and underlying sediments (Brailsford et al. 2019a). This supports the findings set out by Bronk et al. (2007), who suggested that DON can be an extremely important and bioavailable source of N for both bacteria and phytoplankton in aquatic ecosystems.

Nutrient stoichiometry and depletion
In recent years, the influence of stoichiometry of the three major nutrients (C, N, P) has been brought to the forefront of catchment science, with several studies suggesting that modulating nutrient stoichiometry could be help tackle eutrophication in freshwaters (Paerl et al. 2016;Stutter et al. 2018;Rankinen et al. 2019). The influence of nutrient stoichiometry on DOM processing in these catchments has been addressed in previous studies (Brailsford et al. 2019b;Yates et al. 2019b). In these studies and others, C : N ratios of soil have also been found to be a good predictor of both DOC : DON ratios and DOM bioavailability (Kroer 1993;Yates et al. 2019b), while N/P addition has been shown to increase the rate of biodegradation of LMW DOC (Creamer et al. 2014;Brailsford et al. 2019b), as in this study. Reduction of nutrient loading to waters ought, therefore, to lead to a reduction in the rate of consumption of labile and semi-labile DOM in the waterbody by stream heterotrophs, reducing the nutrient resource available to support instream biological production.
In terms of N/P quality, a number of studies have previously demonstrated that although increasing N/P is generally accompanied by increasing DON/DOP concentrations, a decrease in the proportion of DON/DOP components is usually also observed, as the influence of inorganic soluble N and particulate P increases (Perakis and Hedin 2002;Durand et al. 2011;Yates et al. 2019b). In the current study, both DON and DOP degradation were closely related to TN and TP concentrations across our 45 sites (Table 2; Supporting Information Tables S1, S2): the higher the background N and P concentration in the river water, the higher the rate of both DON and DOP depletion. We therefore consider that the increased DON/DOP demand at nutrientenriched sites could be due to: (1) a demand for labile LMW DON/DOP (Bronk et al. 2007) to meet both nutrient resource and energy needs of organisms; (2) all DOM components (DOC, DON, DOP) flushed downstream from upstream sources being utilized by microbial communities for their carbon content meet their energy needs (Battin et al. 2008); (3) changes in the way compounds are metabolized according to differences in nutrient limitation as this varies in both space and time (Brailsford et al. 2019b).

Limitations and future foci
The current study utilized 14 C/ 33 P-labeled tracers to assess the fate of DOC, DON, DOP, and SRP in river waters from 45 sites covering a range of physicochemical gradients. Apart from DON, a single radioisotope tracer was used in each case. Although LMW DOM may represent a fraction of the total OM in aquatic ecosystems, it is a highly diverse mixture, including sugars, amino acids, peptides, organic acids, carboxylic acids, and nucleic acids (Dawson et al. 2001). Further studies are required using mixtures and stoichiometric ratios of isotopically labeled compounds, in order to provide a more representative view of the microbial uptake kinetics of labile LMW DOM in aquatic ecosystems (similar to Brailsford et al. 2019a). In addition, for DOM compounds only the C was radiolabeled, therefore the ultimate fate of N in DON and P in DOP remains unknown.

Conclusions
Comparison of the depletion of DOM components (DOC, DON, DOP) from solution demonstrated that DOC removal from solution proportionally increases with increases in DON and DOP depletion, with DON removal being slightly more rapid and DOP removal being slightly slower than DOC removal, respectively. The depletion of all three DOM fractions increases as nutrient enrichment increases, and as the proportion of HMW DOM decreases along the river enrichment gradient. This study demonstrates that the nutrient chemistry of waters, which is in turn influenced by landcover, can be a predictor of the capacity for DOM processing in the aquatic environment, with inorganic nutrient enrichment and carbon limitation stimulating an increase in DOC, DON, and DOP processing by the microbial biomass. By contrast, microbial SRP processing decreases with increasing nutrient enrichment, suggesting that DOP is preferred over SRP in riverine microbial communities, driving many of the adverse impacts associated with the eutrophication of freshwaters. It is our opinion that downstream "omics" approaches such as primary metabolome analysis could provide more insight into the fate of these compounds following uptake by the aquatic microbial biomass. Programs aiming to control eutrophication in freshwater catchments would therefore need to control both the inorganic and organic N and P loading delivered to waters from their catchments if they are to be effective in realizing improvements in the ecological status of these waters.