Nutrient limitation masks the dissolved organic matter composition effects on bacterial metabolism in unproductive freshwaters

Aquatic microbial responses to changes in the amount and composition of dissolved organic carbon (DOC) are of fundamental ecological and biogeochemical importance. Parallel factor (PARAFAC) analysis of excitation–emission fluorescence spectra is a common tool to characterize DOC, yet its ability to predict bacterial production (BP), bacterial respiration (BR), and bacterial growth efficiency (BGE) vary widely, potentially because inorganic nutrient limitation decouples microbial processes from their dependence on DOC composition. We used 28‐d bioassays with water from 19 lakes, streams, and rivers in northern Sweden to test how much the links between bacterial metabolism and fluorescence PARAFAC components depend on experimental additions of inorganic nutrients. We found a significant interaction effect between nutrient addition and fluorescence on carbon‐specific BP, and weak evidence for influence on BGE by the same interaction (p = 0.1), but no corresponding interaction effect on BR. A practical implication of this interaction was that fluorescence components could explain more than twice as much of the variability in carbon‐specific BP (R2 = 0.90) and BGE (R2 = 0.70) after nitrogen and phosphorus addition, compared with control incubations. Our results suggest that an increased supply of labile DOC relative to ambient phosphorus and nitrogen induces gradually larger degrees of nutrient limitation of BP, which in turn decouple BP and BGE from fluorescence signals. Thus, while fluorescence does contain precise information about the degree to which DOC can support microbial processes, this information may be hidden in field studies due to nutrient limitation of bacterial metabolism.

Bacterial respiration (BR) and bacterial production (BP) are among the largest aquatic carbon transformation processes worldwide (del Giorgio et al. 1997;Jansson et al. 2000), with key roles in the cycling and fate of autochthonous and allochthonous dissolved organic carbon (DOC).When bacteria use DOC for BP, they contribute to biomass production that supports aquatic food chains (Jansson et al. 2007;Creed et al. 2018), whereas BR produces CO 2 that adds substantially to greenhouse gas emissions (Vachon et al. 2017).Given the sensitivity of terrestrial DOC export to a range of ongoing changes in hydrology, land use, and climate (Solomon et al. 2015;Creed et al. 2018), it is crucial to understand how BR and BP in recipient ecosystems respond to increasing DOC of different origin and quality.
Over the past three decades, fluorescence excitation-emission matrix (EEM) analysis has increasingly been used to infer and understand the source (Coble 1996;Walker et al. 2009), bioavailability (Fellman et al. 2010;Graeber et al. 2021), diagenetic state (Wunsch et al. 2018), and chemical composition (Stubbins et al. 2014) of DOC compounds in soils and waters.However, meta-analyses point to a near-ubiquitous distribution of fluorescence peaks in surface waters from around the world (Wunsch et al. 2019), which fits the view that ecosystems with different functionality can show overlap in various optically-derived indices (D'Andrilli et al. 2022).Therefore, the degrees to which fluorescence can be used to predict specific ecosystem processes such as BR and BP remain unclear.
Studies to date report links between peaks in fluorescence EEMs and bacterial metabolism that vary widely in strength (Begum et al. 2023).For example, some studies have shown strong positive correlations between protein-like fluorescence peaks and cumulatively mineralized DOC (through BR) during standard bioassays (Fellman et al. 2008;Pinsonneault et al. 2016), whereas in other studies the links between fluorescence and metabolism have been relatively weak and inconsistent (Berggren and del Giorgio 2015;Berggren et al. 2020).Collectively, this research has not provided an adequate explanation as to why fluorescence varies in its relationship to bacterial metabolism.One hypothesis is that extrinsic environmental factors such as inorganic nutrient supply, which can be a strong co-regulator of bacterial DOC use (Wickland et al. 2012;Mutschlecner et al. 2018;Berggren et al. 2022), dampens or eliminates entirely the connections between metabolism and DOC properties.Therefore, to the extent that fluorescence contains information that can predict BR, BP, and bacterial growth efficiency (BGE), such links may only be visible in experiments where extrinsic factors including nutrients are kept constant and above limitation levels.
We tested the hypothesis that nutrient limitation obfuscates the link between EEMs and bacterial metabolism in laboratory incubation experiments with and without inorganic nitrogen (N) + phosphorus (P) additions, using water from 19 north-Swedish freshwater sites with variable DOC and nutrient concentrations.Microbial metabolism in unproductive boreal freshwaters may in some cases be carbon-limited due to the low bioavailability of the ambient DOC (Soares et al. 2017;Rulli et al. 2022).However, with increases in the concentration of bioavailable DOC, microbial processes should, logically, be increasingly constrained by the supply of inorganic nutrients, which also tends to be low in these remote ecosystems (Myrstener et al. 2022).Therefore, across the broader natural gradients in DOC and nutrients in the regional landscape, we expected that inorganic nutrient limitation of BR, BP, and BGE does occur, and hence fluorescence will explain variability microbial processes better in experiments where the water has been enriched with inorganic macronutrients.

Fieldwork
We sampled 19 freshwater sites in northern Sweden with different physical characteristics on one occasion each in 2018 (Supporting Information Table S1).Seven boreal headwater streams from the Krycklan catchment (Laudon et al. 2013) were sampled during snowmelt in May, four boreal lakes and three larger rivers in the same region as Krycklan were sampled in July during dry climatic conditions, and finally five subarctic lakes were sampled in the far north in Sweden, also during relatively dry conditions (Berggren et al. 2020).Sites were selected to capture the environmental and biogeochemical diversity of this region, with small streams draining bogs, coniferous forests or mixed bog/forest catchments, rivers of different size and water color, and finally small clear-to brown-water lakes spanning two orders of magnitude in catchment sizes (and in the catchment to lake area ratio), thus representing different water residence times.Streams were sampled at V-notch weirs, whereas lakes were sampled at the lake center at 1-m depth from a boat, and larger rivers from flowing water at piers or by wading.See Supporting Information Table S1 for detailed site characteristics.On each occasion, 3.5 L water was obtained in a 1-gal low-density polyethylene Cubitainer ® .Additional water for nutrient analyses was 0.45 μm-filtered into 45-mL Falcon tubes directly in the field and later stored frozen.Samples were kept in cooling bags with ice during transport to the laboratory.

Biodegradation experiments
Upon completion of each sampling campaign (May, July, and August, respectively), biodegradation experiments were immediately initiated to analyze how DOC supported bacterial growth and BR.To remove particulate organic carbon and organisms larger than bacteria, the entire 3.5 L sample from cubitainers were first 1.2 μm filtered through GF/C (Whatman) glass fiber filters.The filtrate from each site was split into two subsamples, where one was used as a control and the other spiked with 0.5 mg N L À1 (35.7 μmol N L À1 ) as NH 4 NO 3 and 50 μg P L À1 (1.61 μmol P L À1 ) as KH 2 PO 4 .The nutrient addition was designed to maintain the ambient mean inorganic N : P molar ratio, which changed slightly from 30 to 23 after treatment while increasing inorganic nutrient concentrations relative to DOC by more than an order of magnitude.The added N and P were more than sufficient to support the cumulative bacterial growth in all incubations, assuming that nutrients were assimilated into biomass at an average "non-limited" molar C : N : P stoichiometry of 86 : 14 : 1 (Godwin and Cotner 2018).
For BR measurements, the control-and nutrient-treated water was transferred to duplicate 500 mL top-filled and gassealed Erlenmeyer flasks, which were incubated in the dark at 20 C for 28 d.We estimated BR from oxygen consumption in each flask, which was measured continuously with Oxy-10 optodes (PreSens).We found no systematic deviations from linearity in oxygen consumption.In fact, there was a nearperfect linear correlation (r 2 = 0.998) between mean oxygen concentration (all samples) and time.Therefore, we estimate BR from the linear slope of dissolved oxygen change over time.To convert BR rates from oxygen units to carbon units, we multiplied the measured molar oxygen consumption rates by a respiratory quotient of 1.2, which is average for unproductive boreal streams and lakes (Berggren et al. 2012) and similar to what has been measured for bacterioplankton in two of our study lakes, Övre Björntjärnen (1.2) and Stortjärnen (1.3), in a previous study (Allesson et al. 2016).
In parallel 500 mL Erlenmeyer flasks (one for each site and treatment), water was kept at 20 C and darkness for measurement of BP.BP was measured on days 1, 3, 7, 14, and 28 using the 3H-leucine incorporation method (Smith and Azam 1992).Triplicate aliquots of 1.2 mL sample water were exposed to 40 nmol L À1 leucine for 1 h, before the incubation was stopped with trichloroacetic acid, and the protein content was purified three times through centrifugation and removal of supernatant.Blanks were pretreated with trichloroacetic acid before the isotope exposure.The uptake of leucine was converted into bacterial carbon (Simon and Azam 1989) by applying the standard conversion factor 1.55 kg C mol Leu À1 multiplied with an assumed intracellular isotopic dilution factor of 2. Total cumulative BP during the incubation (integrated area under the curve) was calculated using the trapezoid rule and transformed into a rate by dividing by the total incubation time.The BGE was calculated as BGE = BP/(BP + BR).

Chemical and optical analyses
Filtered samples for nutrient analyses, representing the field conditions of each sampled site, were shipped frozen to Uppsala University and analyzed at the Department of Ecology and Genetics (EBC).Total dissolved and soluble reactive phosphorus were analyzed through colorimetry using the molybdenum blue method (Menzel and Corwin 1965).For total phosphorus oxidation, potassium persulfate and autoclavation were used.Analysis of NH 4 -N and NO 3 -N was conducted in a Metrohm IC system (883 Basic IC Plus and 919 Autosampler Plus).Total dissolved nitrogen and organic carbon were analyzed on a Shimadzu TNM-L.
Absorbance matrix and EEM were measured on water from the BP incubation flasks, and filtered through GF/F glass fiber filters on days 1, 3, 7, 14, and 28 of the experiment.The scans were collected using an Aqualog (Horiba Scientific) in a 1 cm quartz cuvette (2 s integration time) over 5 nm increments with excitation wavelengths ranging from 230 to 800 nm and emission wavelengths from 250 to 800 nm.The EEMs were blanksubtracted (Milli-Q water), corrected for instrument-specific biases (Cory et al. 2010) and inner filter effects (Kothawala et al. 2013), and normalized to the Raman area of deionized (Milli-Q) water (Lawaetz and Stedmon 2009).Fluorescent components were identified with parallel factor (PARAFAC) analysis of the corrected EEMs using the drEEM toolbox (Murphy et al. 2013).Data was normalized before modeling, and the matrix size was cropped to an excitation range of 250-450 nm and and emission range of 300-600 nm.A four-component model (n = 195) was validated using a sixfold split-half analysis.The model was cross-referenced against the openFLUOR (Murphy et al. 2014) database, showing that all components have been found in previous studies (Table 1).The first three components (C1-C3) have humic-like peaks, whereas component C4 is typical for protein-like or tyrosine fluorescence.Components are shown visually in Supporting Information Fig. S1.
Average total fluorescence intensity did not differ by more than 6% between any time points (this was also similar for all individual components; not shown), hence we considered that fluorescence was approximately constant during the incubations.We, therefore, calculated integrated average fluorescence intensities during the incubation for each sample by using the trapezoid rule (area under the curve) and dividing by the number of integrated days.

Statistical analyses
To test the interaction effect between nutrient addition and fluorescence on bacterial metabolism, we first needed to decrease the number of explanatory fluorescence variables to retain sufficient degrees of freedom given our limited sample size.Therefore, we used ordinary least square regression to combine the different fluorescence components in all samples across treatments (n = 38) into one single (regressed) variable for each aspect of metabolism.Absolute BP and BR were regressed from linear combinations of absolute fluorescence components in Raman units using the call "ols_regress" in package "olsrr" for R 3.5.1.(R Development Core Team 2018).In the same way, the relative response variables, that is, BGE and carbon-specific BP and BR (BP DOC À1 and BR DOC À1 ) were regressed from linear combinations of relative Protein-like, tyrosine fluorescence components (% of total fluorescence).For each aspect of metabolism, we then again used "ols_regress" to generate a new set of regressions, where standardized coefficients were derived for effects of nutrient addition (dummy variable; 0 or 1) and fluorescence (the linearly regressed variables), in models with and without an interaction term (Nutr*Fluo).
We then explored relationships between bacterial metabolism variables and individual PARAFAC components, using linear regression and Pearson r correlation coefficients (tested two-tailed) in the R package "olsrr".For response variables affected by significant (p < 0.05) or near-significant (p $ 0.1) interaction effects between nutrients and fluorescence, we tested relationships with individual PARAFAC components separately in control and nutrient-amended samples, respectively, to test how nutrient addition in practice improved the ability to predict bacterial metabolism using DOC composition data.Again, the absolute rate variables BR and BP were explained by means of absolute fluorescence intensities in Raman area units, while the relative response variables BR DOC À1 , BP DOC À1 , and BGE were explained by relative fluorescence components in units of percentage.
Finally, we used molar ratios between labile C concentrations and different nutrient variables (N and P concentrations) to explain the degree of nutrient limitation in bacterial growth.To calculate these ratios, we assessed the potentially labile C as the sum of the respired DOC in nutrient-amended incubations.The degree of nutrient limitation was considered as the BP rate in nutrient-amended incubations minus the corresponding control incubation rate.
Given the arbitrary nature of normality tests, we inspected the Gaussian distribution of variables and model residuals manually.We assumed normality in all cases except for BP per unit of DOC, which was strongly skewed (Fisher's skewness > 2) and produced asymmetric model output.This problem was solved by log10-transformation of BP DOC À1 before any statistical analysis.

Results
Initial DOC concentrations spanned a gradient from 1.5 to 14.7 mg L À1 across the samples (Supporting Information Table S3).BP, dissolved oxygen concentrations, and total fluorescence intensity varied broadly among samples during the 28-d dark incubations (Fig. 1).Based on these data, we derived average time-integrated values of BR, BP, and fluorescence component intensities for each sample (Supporting Information Table S2), which provide information about how bacteria used the DOC.
In regression models without interaction terms (Fig. 2), the dummy variable for nutrient addition had significant    S4).On average, the cumulative BR during the incubations mineralized 8.8% AE 3.4% of the DOC in the control incubations and 14.2% AE 4.4% of the DOC after nutrient additions (mean AE SD).Also, BP was strongly nutrient-limited, with average values twice as high after nutrient addition (30.1 AE 10.0 μg C L À1 d À1 ) relative to controls (14.4AE 9.7 μg C L À1 d À1 ), but there was no correlation between BP and DOC.Nutrient addition was not a significant variable in the BGE regression models (Fig. 3).Mean BGE was 37.8% AE 16.7% in the control incubations, but not clearly different (47.0 AE 18.2%) when nutrients were added.BGE was negatively related to DOC concentration in nutrient-amended incubations (Pearson r = À0.81,n = 19, p < 0.0001), but this relationship was considerably weaker in control incubations (r = À0.45,n = 19, p = 0.051).
All bacterial metabolism response variables except BR per unit DOC were significantly related to fluorescence (Figs. 2, 3; models without interaction term).In absolute units, BR was strongly positively related to fluorescence components C1-C3, and these components were in turn strongly positively related to DOC concentrations (Supporting Information Table S4).However, BR per unit DOC showed no relationship with any fluorescence component.BP and BGE both increased with C2 (microbial-like) and C4 (protein-like) components and decreased with terrestrial humic-like components C1 and C3 (Supporting Information Table S4).Finally, we found no significant interaction effect between nutrient addition and fluorescence for absolute BR or BP (Fig. 2).However, for log(BP DOC À1 ) the interaction coefficient was significantly positive (p = 0.027; Fig. 3), meaning that fluorescence regulated the BP differently after nutrient addition, compared to in control incubations.For BGE and BR DOC À1 , the magnitude of the interaction was high and positive, but at the same time the interaction coefficient had high variability (error), hence it was not significant for BR DOC À1 (p = 0.312) and was only marginally close to significant ( p = 0.122) for BGE (Fig. 3).
Given the significant or near-significant interaction effect between fluorescence and nutrients found for log(BP DOC À1 ) and BGE, we examined their relationships with individual fluorescence PARAFAC components separately in nutrient-amended incubations and controls.Overall, these relationships were considerably stronger after nutrient addition (Fig. 4).This was true for all fluorescence components, but it can be noted that C1 alone explained as much as 71% and 90%, respectively, of the variance in BGE and BP per unit DOC in nutrient-amended incubation, whereas corresponding values in control incubations were 28% and 37% (see equations in Fig. 4).When trying all linear combinations of fluorescence variables to explain the response without nutrient additions, the highest possible adjusted R 2 was still only 0.29 for BGE (À3.55ÁC1 % + 2.57ÁC3 % + 153.48) and 0.44 for log10-transformed BP DOC À1 (À0.06ÁC1 % + 0.10ÁC3 % + 0.05ÁC2 % À 1.28).For nutrient-amended samples, no combination of fluorescence variables gave higher adjusted R 2 than regressions based on C1 alone, that is, 0.71 for BGE and 0.90 for log(BP DOC À1 ).
The total amount of potentially labile C in sampled waters, assessed as the sum of the respired DOC in nutrient-amended incubations, varied from 13.4 to 250.8 μmol L À1 , with a mean AE SD of 90.1 AE 54.5 μmol L À1 .The response in BP to nutrient addition was strongly positively related to the ratio between this labile C pool and ambient dissolved inorganic N (Fig. 5a; r 2 = 0.78, n = 19, two-tail p < 0.001), but also to the ratio between labile C and total P (Fig. 5b; r 2 = 0.60, n = 19, two-tail p < 0.001).We were not able to explain nutrient limitation in any microbial parameter by using element ratios with total DOC concentration as the numerator.

Discussion
There has been extensive research on the chemical composition of natural dissolved organic matter during the past decade (McCallister et al. 2018), but surprisingly little is known about how DOC composition affects the fate of DOC Fig. 3. Standardized coefficients (symbols AE standard error) from linear regressions of carbon-specific rates of (a) bacterial respiration and (b) bacterial production, and (c) bacterial growth efficiency, explained in models to the left (filled symbols) by nutrient additions ("Nutr"; dummy variable) and fluorescence ("Fluo"), and in models to the right (filled symbols) by Nutr, Fluo plus an interaction term ("Nutr*Fluo").Here, Fluo is a linear combination, derived from a separate regression, of four PARAFAC components in units of percentage.
in practice (Berggren et al. 2022).Our study points to two major reasons why DOC bioreactivity predictions based on organic matter fluorescence EEM analysis with PARAFAC might have failed in past studies.First, we show that DOC composition revealed by EEM analysis is a stronger regulator of bacterial growth than respiration, suggesting that fluorescence is of limited use for the prediction of DOC mineralization alone.Second, the nutrient limitation can decouple the link between DOC composition and bacterial metabolism in unproductive freshwaters, again limiting the utility of fluorescence in studies of bacterial metabolism in nature.Nonetheless, given a sufficient supply of N and P, the fluorescence EEMs contain valuable information about the degree to which DOC supports BP, and about the BGE when using the DOC.
In terms of the overall span of BP, BR, and BGE in this study, our values align well with what has been reported in similar incubation studies.The percentage of respired DOC was low compared with that in aquatic incubation studies reviewed by del Giorgio and Davis ( 2003), but similar to or higher than reported 1-month values for Swedish unproductive clear-and brown-water lakes in Koehler et al. (2012).Interestingly, whereas BR per unit DOC was relatively constant, the BP per unit DOC showed large variability, reflected in a strong BGE gradient across the sites.Compared with other studies, the mean BGEs of 38% and 47% in our control and nutrientamended incubations, respectively, suggest a relatively high potential for the metabolized DOC to support BP (del Giorgio and Cole 1998).However, absolute values of BP derived from leucine incorporation rates should be interpreted with some caution, due to variations in leucine-to-C conversion factors between studies (Kirchman 1993;Calvo-Diaz and Moran 2009).Thus, whereas BR did not stand out as particularly high in this study, BP and BGE were in the upper range of what could be expected, yet within reasonable limits.
Rates of BR were constant during our 28-d incubations, which is surprising given that DOC mineralization in dark incubations normally is relatively faster in the beginning (Guillemette and del Giorgio 2011;Koehler et al. 2012).It is similarly surprising that the fluorescence intensity was constant, which contrasts with previous findings of dynamic production and consumption of fluorescent molecules during microbial transformations of labile DOC (Guillemette and del Giorgio 2012;Berggren et al. 2020).A possible explanation for these patterns is the drought in northern Europe during the study year 2018 (Bakke et al. 2020), which affected our July and August samples (site nos. 8-19).Dry conditions decrease the hydrological connectivity between superficial organic forest soils and streams (Laudon et al. 2011;Gomez-Gener et al. 2020), which results in DOC from deeper soil sources with lower biological reactivity (Buffam et al. 2001;Wilson et al. 2016).Moreover, as our stream samples from May were obtained in the later part of the spring flood period (site nos.1-7), most of the labile DOC had probably already been flushed from the catchment soils (Wilson et al. 2013) before sampling.Thus, there was little carbon available to support especially high BR at the beginning of the incubations (Guillemette and del Giorgio 2011) and, therefore, DOC composition and reactivity were relatively constant during our incubations.
In past studies, the degree to which fluorescence can explain variation in BR has been highly variable, but many papers report positive correlations between % protein-like fluorescence and % mineralized or respired DOC (Fellman et al. 2008;Hosen et al. 2014;Pinsonneault et al. 2016;Berggren et al. 2020).In this study, BR in absolute units was positively correlated to fluorescence intensities from components C1-C3, but these correlations were likely confounded by strong underlying control of both BR and fluorescence by DOC concentration.Indeed, there was no correlation between BR per unit of DOC and any fluorescence component, and thus fluorescence could not explain any additional variability in BR besides that captured by DOC concentration.Thus, BR per unit DOC was more or less invariant in both control and nutrient-amended incubations, similar to what has been reported before in the Krycklan catchment (Berggren et al. 2007), but different from past measurements in larger Swedish rivers (Soares et al. 2019) and lakes in the Abisko region (Berggren et al. 2020).In this study, the variability in DOC source lability was apparently not large enough to create gradients in BR per unit of DOC.
In agreement with our hypothesis, the variables which were significantly or near-significantly affected by an interaction between nutrient addition and fluorescence, that is, BP per unit DOC and BGE, showed systematically stronger links to fluorescence in nutrient-amended than in control incubations.Past studies have found indications of N (Berggren et al. 2007) and P (Soares et al. 2019) limitation of bacterial metabolism in some of the same study sites included here, consistent with widespread observations of nutrient limitation in aquatic ecosystems across this broader region (Myrstener et al. 2022).Therefore, we recommend that both N and P are added (combined) in studies that aim at identifying links between dissolved organic matter quality and bacterial metabolism.Given adequate N and P supply, our results show that the relative percentages of different fluorescence components (i.e., the composition) give valuable information about the growth variables BGE and BP per unit DOC.
Since our experiment added N and P only in combination, we are not able to test when and to what extent each of the two nutrients limited bacterial metabolism.More generally, in unproductive, organic-rich waters, it is difficult to predict the limiting element based on bulk element stoichiometry, due to the dominance of organic nutrients that can have highly variable bioavailable fractions (Berggren et al. 2015;Soares et al. 2017;Rulli et al. 2022).Nonetheless, we found that the response in BP to nutrient addition could be explained by the ratios between labile DOC (amount respired), inorganic N, and TP (as shown in Fig. 5).These ratios (Fig. 5) were typically several times higher than those in bacterial biomass (C : N = 14  S3).
and C : P = 204) during N + P co-limiting conditions (Graeber et al. 2021), which may suggest that N and P both limited the bacterial metabolism.This possibility holds true even when considering that biomass ratios underestimate the relative C supply need of bacteria by a factor that is inversely proportional to BGE (del Giorgio and Cole 1998).Moreover, previous studies in northern Sweden have reported both N (Berggren et al. 2007) and P (Jansson et al. 1996) limitations of bacterioplankton.Still, it is not possible to draw specific conclusions regarding nutrient status in our experiments, because unknown and highly variable fractions of organic P (Jansson et al. 2012) and organic N (Soares et al. 2017) likely contributed to the true supply ratios of bioavailable nutrients.
The fact that fluorescence data explained bacterial metabolism much better in nutrient-amended than in control incubations sheds new light on the discussion regarding causality in fluorescence-metabolism relationships (Berggren et al. 2020).In practice, fluorescence has often been thought to reflect reactive DOC constituents (Fellman et al. 2008;Lapierre and del Giorgio 2014) that directly promote microbial metabolism.However, several fluorescence components, including proteinlike, are produced because of the bacterial metabolism of reactive DOC (Guillemette and del Giorgio 2012;Fox et al. 2017;Berggren et al. 2020); hence, the causality might be reversed.Nonetheless, our study points to a potentially hidden and causal link from fluorescent DOC sources to bacterial growth because the relationships became stronger after nutrient amendment, which is when C quality would be expected to regulate metabolism.This underscores that fluorescence indeed can contain information about potential DOC support of microbial processes, but this information may not be useful in field studies due to nutrient limitation of bacterial metabolism.
Nutrient limitation in freshwater bacterioplankton processes has been relatively less researched compared with that in phytoplankton processes, but both are utterly affected by bioavailable N, P, and organic C fractions (Jansson et al. 1996).As such, the influences of DOC composition on bacterioplankton metabolism can be hypothesized to follow spatial and temporal trends in environmental changes that alter stream, river, and lake chemistry (Isles et al. 2018;Isles et al. 2021).Across much of Fennoscandia, the most important of such changes include water "browning", that is, increased DOC loading from land and concurrent reductions in inorganic N and P supply (Lucas et al. 2016;Huser et al. 2018;Mason et al. 2022).Given apparent trends of increasing DOC-to-nutrient ratios in this region (Mosquera et al. 2022), our results suggest that the influences of DOC composition (such as that revealed by fluorescence) on bacterial metabolism are likely to weaken and be increasingly replaced by the constraints imposed by nutrient limitation.

Fig. 1 .
Fig. 1.Distribution of (a) bacterial production, (b) dissolved oxygen, and (c) total fluorescence intensity values in water from 19 freshwater sites, measured during different discrete time points T (day 1, 3, 7, 14, and 28) of 28-d dark incubations in gas-tight bottles.White boxes represent control samples and gray boxes samples amended with inorganic nitrogen and phosphorus.Boxes present quartiles 1, 2 (median), and 3. Lower and upper whiskers reach toward the 5 th and 95 th percentiles, respectively, and caps show mean/max values.

Fig. 2 .
Fig. 2. Standardized coefficients (symbols AE standard error) from linear regressions of absolute rates of (a) bacterial respiration and (b) bacterial production, explained in models to the left (filled symbols) by nutrient additions ("Nutr") and fluorescence ("Fluo"), and in models to the right (filled symbols) by Nutr, Fluo, and their interactions ("Nutr*Fluo").Nutr is a dummy variable for nutrient enrichment (0 = no addition; 1 = nitrogen plus phosphorus addition) and Fluo is a linear combination, derived from a separate regression, of four fluorescence PARAFAC components in Raman area units.

Fig. 5 .
Fig. 5. Increase in average bacterial production in response to nutrient (NP addition) plotted against (a) the ratio between labile C and dissolved inorganic nitrogen (DIN), where DIN is the sum of nitrogen as NH 4 , NO 3 , and NO 2 , and (b) the ratio between labile C and ambient total phosphorus.In the lower (b) figure, TP is used in the x-axis ratio instead of soluble reactive phosphorus, because the latter had a concentration close to 0 in several samples (Supporting Information TableS3).