Stable carbon isotope biogeochemistry of lakes along a trophic gradient

16 The


Introduction
Studies suggest that lakes contribute significantly to the global carbon budget via organic matter burial and emission of CO 2 to the atmosphere (Cole et al., 2007).The balance between primary production and external organic carbon input on the one hand and respiration and burial of organic carbon on the other governs whether individual lakes are sources or sinks of CO 2 .
This metabolic balance can be disturbed by changes in nutrient or organic matter inputs to the lake.Primary (autochthonous) production increases with increasing nutrient concentrations and lakes with high autochthonous carbon production, i.e. eutrophic lakes, may be sinks for CO 2 (Schindler et al., 1997).The loading of allochthonous (terrestrial) carbon is a key factor controlling community respiration of lakes.The metabolic balance of lakes is directly influenced by allochthonous organic carbon loading and trophic state (Del Giorgio and Peters, 1994;Hanson et al., 2003).
Stable carbon isotope analysis is a powerful tool for studying carbon cycling in lakes since it allows studying inorganic and organic carbon pools and changes therein.It can provide information on the metabolic balance and the sources of organic matter fueling respiration.
Respiration yields 13 C-depleted carbon dioxide from organic matter with the result that δ 13 C of dissolved inorganic carbon of the lakes becomes lower (Parker et al., 2010).Primary producers preferentially incorporate 12 C in their organic matter with the consequence that the remaining pool of dissolved inorganic carbon will be enriched in 13 C (Herczeg, 1987;Parker et al., 2010).
The δ 13 C of the dissolved inorganic carbon pool thus integrates the relative importance of respiration and primary production (Parker et al., 2010).The δ 13 C of organic carbon pools is primarily governed by the δ 13 C of the dissolved inorganic carbon used by primary producers and the isotope fractionation during carbon fixation.Terrestrial plants use atmospheric carbon dioxide while aquatic primary producers utilize dissolved carbon dioxide or bicarbonate (Fry, 2006).The δ 13 C of terrestrial-derived organic carbon is therefore often distinct from that of organic matter produced within the lakes and this difference can be used to trace carbon flows and origins in plankton food webs.
A major challenge in stable isotope studies is to elucidate the isotopic composition of microbial organisms (Middelburg, 2014), such as phytoplankton and bacteria, since it is difficult to separate these potential carbon sources from bulk particulate organic carbon (POC).
Therefore, most studies use indirect methods to determine δ 13 C of phytoplankton (δ 13 C phyto ), bacteria (δ 13 C bac ) and allochthonous carbon (δ 13 C allo ).Common methods for determining δ 13 C phyto are the use of the δ 13 C of particulate organic carbon (POC) with correction for nonphytoplankton carbon and estimates based on δ 13 C of dissolved inorganic carbon (DIC) with an isotope fractionation factor (ε), obtained from experimental studies.Other methods are the use of zooplankton consumers as a proxy for δ 13 C phyto or size fractionation of organic matter and subsequent determination of δ 13 C of different size classes (Marty and Planas, 2008).
Isotopic ratios of bacteria in field studies have been derived from re-growing bacteria in bioassays (Coffin et al., 1989) or dialysis cultures (Kritzberg et al., 2004), with measurement of 13 C in POC or respired CO 2 (McCallister et al., 2008) and from biomarkers like nucleic acids (Coffin et al., 1990) and lipids (Bontes et al., 2006;Pace et al., 2007).Some studies used δ 13 C of DOC as proxy for δ 13 C of bacteria, assuming that DOC was the primary carbon source for bacteria (Taipale et al., 2008;Zigah et al., 2012).
A commonly used proxy for allochthonous δ 13 C is the δ 13 C of terrestrial C 3 plants, which dominates most terrestrial vegetation and has a δ 13 C of ~-28 ‰ (Fry, 2006;Kohn, 2010).When vegetation is dominated by C 4 plants, however, common in tropical areas and agricultural areas with maize production (δ 13 C of ~-14‰; Fry, 2006), the isotopic composition of allochthonous carbon can be significantly enriched in 13 C.In lakes with large terrestrial input, δ 13 C of DOC can be used as a proxy for δ 13 C allo , since terrestrial carbon forms the largest fraction of DOC (Kritzberg et al., 2004;Wilkinson et al., 2013).
Compound specific isotope analysis (CSIA) of polar lipid-derived fatty acids (PLFA) biomarkers has shown to be a valuable tool to determine the isotopic composition of plankton producers and consumers (Boschker and Middelburg, 2002).Groups of phytoplankton and bacteria have different fatty acid (FA) compositions, so by analyzing the δ 13 C of specific FA, the δ 13 C of phytoplankton and bacteria can be inferred.The combined use of stable isotopes and FA biomarkers has been successfully applied to study autochthonous and allochthonous carbon contributions to zooplankton in a tidal river (Van den Meersche et al., 2009).Few studies have applied CSIA to study carbon flows in plankton food webs in lakes.Examples are a phytoplankton-zooplankton interaction study in a eutrophic lake (Pel et al., 2003), a biomanipulation effect study (Bontes et al., 2006), a 13 C lake enrichment study (Pace et al., 2007), and a cyanobacteria-zooplankton interaction study (de Kluijver et al., 2012).
In this study, we used compound-specific isotope analyses to examine carbon flows in plankton food webs in temperate (North American) lakes.The lake survey encompassed a range in trophic states from oligotrophic lakes, with an expected dominance of allochthonous input, to eutrophic lakes, with an expected lower allochthonous input.In this trophic range, we explored patterns of isotopic variability in dissolved inorganic and organic carbon, particulate organic carbon and carbohydrates, phytoplankton, allochthonous carbon, heterotrophic bacteria and their relationships.

Site description
The 22 lakes sampled in this study are located in Iowa and Itasca County in Minnesota, USA.Iowa lakes are mostly man-made and situated in a highly agricultural region, with maize and soya beans as main products.This type of row-crop agriculture has a large impact on the nutrient load of the lake watershed (Arbuckle and Downing, 2001).However, Itasca lakes are natural and situated in a highly forested area.The catchment areas have developed since the last glaciation episode ca.12,000 years ago and consist of carbonate-poor glacial deposits (till) (Grimley, 2000).

Field sampling
The lakes were sampled in July -August 2009 as part of the ongoing lake monitoring program of the limnology laboratories of Iowa State University and Itasca Community College.
Key parameters, such as temperature, pH, Secchi transparency, oxygen, inorganic nutrients and carbon concentrations were measured as part of and according to the lake monitoring program.
All samples were taken from up to 2 m of the upper mixed zone at the deepest point of each lake.
Water samples were taken between 10h-16h, a period of the day that yields relatively stable water chemistry readings in these lakes.More information on data collection, lake characteristics, and methods can be found on http://limnoweb.eeob.iastate.edu/itascalakesand http://limnology.eeob.iastate.edu/lakereport.All nutrients were analyzed using certified methods and strict quality assurance procedures.
Triplicate water samples were taken for stable isotope analyses and concentrations of the major carbon pools.Headspace vials (20 ml and 2 ml) were filled on board with sampled water using the overflow method and sealed with gas-tight caps for DIC isotope analyses and concentrations, respectively.Mercury chloride was added for preservation and the samples were stored upside-down at room temperature.For DOC analyses, 20 ml of sampled water was filtered over GF/F (0.7 µm pore size, 25 mm diameter) and stored frozen in clean (acid and milli-Q rinsed) vials until further analysis.
Seston samples for particulate organic carbon (POC) and carbohydrates were collected by filtering 0.4 to 1 L of sampled water on pre-weighed and pre-combusted GF/F filters (0.7 µm pore size, 47 mm diameter), which were subsequently dried at 60° for POC analysis or freezedried for carbohydrates; PLFA samples were collected by filtering ~2 L sampled water on precombusted GF/F filters (0.7 µm, 47 mm) and filters were stored frozen.Pigment samples were taken for concentrations only and collected by filtering ~600 ml sampled water on GF/F filters (0.7 µm, 47 mm) in the dark and filters were stored frozen.

Laboratory analyses
POC samples were analyzed for carbon content and isotope ratios on a Thermo Electron Flash EA 1112 analyzer (EA) coupled to a Delta V isotope ratio mass spectrometer (IRMS) (c.f.Nieuwenhuize et al., 1994).For DIC isotope analyses, a helium headspace was created in the headspace vials and samples were acidified with H 3 PO 4 solution.After equilibration, the CO 2 concentration and isotope ratio in the headspace was measured using EA-IRMS (Gilikin and Bouillon, 2007).DIC concentrations were measured using spectrophotometry according to Stoll et al., (2001).For DOC analyses, the samples were acidified and flushed with helium to remove DIC and subsequently oxidized with sodium persulfate (Na 2 S 2 O 8 ); the isotope ratio and concentration of CO 2 resulting from this treatment was measured using high performance liquid chromatography -isotope ratio mass spectrometry (HPLC-IRMS) (Boschker et al. 2008).PLFA samples were extracted according to a modified Bligh and Dyer method (Bligh andDyer, 1959, Middelburg et al., 2000).The lipids were fractionated in different polarity classes by column separation on a heat-activated silic acid column and subsequent elution with chloroform, acetone, and methanol.The methanol fractions, containing most of the PLFA, were collected and transformed to fatty acid methyl esters (FAME) using methanolic NaOH.The 12:0 and 19:0 FAME were added as internal standards.Concentrations and δ 13 C of individual PLFA were measured using gas chromatography-combustion isotope ratio mass spectrometry (GC-C-IRMS) (Middelburg et al., 2000).The isotopic compositions were corrected for the carbon added during derivatization.Pigment samples were extracted with 90% acetone in purified (miliQ) water with intense shaking.Concentrations of individual pigments were measured on HPLC (Wright et al. 1991).Carbohydrate samples were hydrolyzed in H 2 SO 4 , neutralized with SrCO 3 , and precipitated with BaSO 4 .The supernatant was collected and measured using HPLC-IRMS according to Boschker et al., (2008).

Data analyses
The lakes were divided into eutrophic and oligo-mesotrophic lakes based on average summer total phosphorus (TP) concentrations.Lakes with TP values >24 µg L -1 and a corresponding trophic state index >50 were classified as eutrophic, and lakes with TP values < 24 µg L -1 as oligo-mesotrophic (Carlson 1977).All lakes in Iowa and one lake in Minnesota were classified as eutrophic, while all oligo-mesotrophic lakes were located in Minnesota.

CO 2 system and isotopic composition of CO 2
The different components of the CO 2 system were calculated from temperature, laboratory pH, and DIC concentrations using a salinity of 0 using the R package AquaEnv (Hofmann et al., 2010).Stable isotope ratios were expressed in the delta notation (δ 13 C), which is the 13 C/ 12 C ratio relative to VPDB standard, in part per thousand (‰).The isotope ratio of CO 2 (aq) was calculated from δ 13 C DIC according to Zhang et al., (1995): CO in total DIC, calculated from pH and DIC concentrations.

δ 13 C of phytoplankton and bacteria
Poly-unsaturated fatty acids (PUFA) are abundant in most phytoplankton, and can generally be used as chemotaxonomic markers for this group (Dijkman and Kromkamp, 2006).

δ 13 C of allochthonous carbon
Allochthonous organic carbon (δ 13 C allo ), i.e. organic matter delivered to lakes as DOC or POC, cannot be measured directly as such and we therefore used two proxies: the measured isotopic ratios of DOC (δ 13 C DOC ) and calculated isotopic composition of particulate detritus (δ 13 C det ).The latter was calculated from a mass balance and mixing model, similar to Marty and Planas, (2008), amended with zooplankton and bacteria.We assumed that POC consists of phytoplankton, detritus, bacteria, and zooplankton, and that the δ 13 C of POC represents a mixture of the weighted δ 13 C of the individual groups.Subsequently, δ 13 C det in each lake was derived from δ 13 C POC : The latter equation simply states that detrital organic matter is the non-living part of total particulate organic matter pool.
Phytoplankton carbon (C phyto ) (mg C L -1 ) was calculated as the average of biomass estimates based on chl a concentration (C: Chl a = 50) as well as phytoplankton FA derived biomass, to minimize the error associated with each method.Phytoplankton FA biomass was calculated from the sum of phytoplankton PLFA (∑ 18:3ω3, 18:4ω3, 20:5ω3, 22:6ω3, and 20:4ω6) and a C: specific FA ratio of 60 based on culture studies, summarized in Dijkman and Kromkamp (2006).The two approaches yielded similar results.Bacterial carbon (C bac ) (mg C L - 1 ) was calculated from the summed concentrations of bacteria specific FA (i15:0, ai15:0, and i17:0) and a C bac : FA ratio of 50 (Middelburg et al., 2000).Zooplankton carbon (C zoo ) used in equation 3 was estimated to be ~10 % of C phyto (Del Giorgio and Gasol, 1995) and zooplankton δ 13 C are based on de Kluijver (2012).Uncertainties in δ 13 C and biomass of phytoplankton, bacteria and zooplankton were not considered in calculating δ 13 C det .

Statistical analyses
Data that were part of the lake monitoring program and pCO 2 values represent single samples of each lake.Data on carbon concentrations and isotopic compositions in each lake convey averages of triplicate samples.Statistical analyses were done with software package "R" (R development core team, 2013).Prior to correlation analyses, data were checked for normal distribution (Shapiro test) and log-transformed when necessary to achieve normal distribution.
Correlation coefficients were calculated using Pearson product-moment correlation coefficient (normal distribution) or Spearman's rank correlation coefficient (non-normal distribution).For reasons of instructiveness we present the average ± sd values for eutrophic lakes (n=11) and oligo-mesotrophic lakes (n=11), but we do realize that any division based on a concentration is somewhat arbitrary.The correlations were tested for total lakes (n=22) and for eutrophic lakes (n=11) and oligo-mesotrophic lakes (n=11).Differences between eutrophic and oligomesotrophic lakes were statistically tested using Student's t-tests.

Lake chemistry
The sampled lakes covered a large range of nutrients and CO 2 system characteristics (Table 1).DIC values ranged from 0.05 to 4.55 mmol L -1 , alkalinity values ranged from 0.070 to 2.4 mmol L -1 and pH ranged from 6.1 to 9.8 (Table 1).The calculated pCO 2 values were in the range from 10-4500 µatm, covering a broad range from under-to super-saturation.The CO 2 system in eutrophic and oligo-mesotrophic lakes was clearly different.On average, the eutrophic lakes had higher DIC, alkalinity, and pH than the oligo-mesotrophic lakes (Fig. 1, Table 1).In the eutrophic lakes, there were positive correlations between alkalinity and DIC and pCO 2 values (Fig. 1A, Table 2) and a negative correlation between pH and pCO 2 (Fig. 1B, Table 2).The pCO 2 values were not related to pH, alkalinity, or DIC in the oligo-mesotrophic lakes.Both lake systems showed super-saturation (average pCO 2 838 µatm in both), but the pCO 2 range was much larger in eutrophic lakes (10-4500 µatm) compared to oligo-mesotrophic lakes (310-3200 µatm) (Fig. 1, Table 1).

Organic carbon and fatty acid concentrations
POC, C phyto and C bac concentrations were higher and DOC concentrations were lower in the eutrophic lakes compared to the oligo-mesotrophic lakes (Table 1).C phyto was on average 1.32 ± 1.10 mg C L -1 and 0.11 ± 0.03 mg C L -1 , corresponding to 44 % ± 17 % and 10 % ± 5 % of POC in eutrophic and oligo-mesotrophic lakes, respectively.C phyto and C bac were significantly related to TP (Table 2), but not to total nitrogen (TN) concentration.Average C bac was 0.114 ± 0.081 mg C L -1 and 0.021 ± 0.017 mg C L -1 in eutrophic and oligo-mesotrophic lakes, respectively.
Overall, lake pCO 2 decreased with increasing C phyto , but the effect was strongest in eutrophic lakes (Fig. 1C, Table 2).In the oligo-mesotrophic lakes, pCO 2 increased with increasing DOC (Fig. 1D, Table 2), but this effect was caused by one point: the high pCO 2 at high DOC in lake Sturgeon.In the eutrophic lakes, DOC concentrations were lower compared to the oligo-mesotrophic lakes and did not act on lake pCO 2 (Fig. 1D, Table 1).

δ 13 C of DIC and CO 2
The isotope ratios of the major carbon pools in each lake are presented in Table 3 and in boxplots (median and percentiles) in Figure 2. δ 13 C DIC ranged from -9.3 to +1.5 ‰ and 2 CO 13 C  (derived from δ 13 C DIC ) was on average 10.9 ± 0.3 ‰ depleted in 13 C relative to DIC, with a range of -20.8 to -8.9 ‰. δ 13 C DIC and 2 CO 13 C  showed no correlation with alkalinity, DIC, pH, temperature and lake area.A weak negative relation between pCO 2 and δ 13 C DIC was observed, which was stronger in oligo-mesotrophic lakes than in eutrophic lakes (Fig. 3A, Table 2).The highest pCO 2 lakes had the lowest δ 13 C DIC , suggesting that respiration of organic matter influenced δ 13 C DIC .Low CO 2 lakes had enriched δ 13 C DIC , indicating influence of primary production.Weak, but significant relations were observed for POC and DOC with δ 13 C DIC (Table 2).In eutrophic lakes, δ 13 C DIC increased with increasing POC (Fig. 3B, Table 2), while in oligomesotrophic lakes, δ 13 C DIC decreased with increasing DOC (Fig. 3C, Table 2).

δ 13 C of organic carbon pools
The isotopic composition of DOC (δ 13 C DOC ) had the narrowest range of all carbon pools, only -28.8 to -27.0 ‰ (mean -28.0 ‰) in the oligo-mesotrophic lakes and a slightly larger range of -27.6 to -23.7 ‰ (mean -25.4 ‰) in the eutrophic lakes (Fig. 2, Table 3).The δ 13 C range of POC (δ 13 C POC ) was larger than of DOC in both lake types and on average 2.0 ‰ lower compared to δ 13 C DOC , with mean values of -27.8 ± 3.6 ‰ in eutrophic and -29.7 ± 2.8 ‰ in oligomesotrophic lakes (Fig. 2, Table 3).δ 13 C of particulate glucose (δ 13 C gluc ), the most abundant carbohydrate, was always enriched in 13 C compared to δ 13 C POC and the enrichment was similar in eutrophic (3.1 ± 1.7 ‰) and oligo-mesotrophic lakes (2.8 ± 1.5 ‰) (Fig. 2).On the contrary, the concentration-weighted average δ 13 C of all fatty acids increased with increasing TP (Table 2).

Carbon isotopic composition of phytoplankton
δ 13 C phyto depends on the isotopic composition of substrate ( 2 CO 13 C  ), the isotope fractionation ) (ε phyto CO 2  associated with primary production and the isotopic difference between PUFA's and biomass.δ 13 C phyto in the eutrophic lakes became more enriched in 13 C with increasing C phyto (Fig. 5A, Table 2) and decreasing pCO 2 (Fig. 5B, Table 2).No relation between δ 13 C phyto and C phyto was observed in the oligo-mesotrophic lakes (Fig. 5A), but there was a strong negative relation with pCO 2 (Fig. 5B, Table 2).The influence of C phyto on δ 13 C phyto in the eutrophic lakes was also reflected in fractionation; phyto CO 2 ε  was highly variable in eutrophic lakes, while it was less variable in oligotrophic lakes (Fig. 5C).The range of phyto CO 2 ε  was 7.8 to 24.7 ‰ (mean 16.9 ‰) in eutrophic and 11.7 to 18.8 ‰ (mean 17.1 ‰) in oligo-mesotrophic lakes, when δ 13 C phyto_cor was used (Table 3).The less variable ε in oligo-mesotrophic lakes resulted in a strong correlation between 2 CO 13 C δ and δ 13 C phyto (Table 2), which was absent in the eutrophic lakes.phyto CO 2 ε  correlated negatively with C phyto in eutrophic lakes, however (Table 2).
The variability in δ 13 C phyto in eutrophic lakes can be mainly attributed to the presence of two clusters: a 13 C-enriched cluster at the highest C phyto and a 13 C-depleted cluster at lower C phyto (Fig. 5A).Interestingly, the eutrophic lakes within the 13 C-enriched cluster also had high concentrations of zeaxanthin, a marker pigment for cyanobacteria (data not shown here).

Lake metabolism, pCO 2 and δ 13 C DIC
In our study, about 3/4 of the lakes were supersaturated with pCO 2 , consistent with the literature that lakes generally emit carbon dioxide to the atmosphere (Cole et al., 1994, Cole et al., 2007).This CO 2 excess can be due to in-lake respiration of terrestrially derived organic carbon outbalancing carbon dioxide fixation by primary producers (negative metabolic balance) or due to river and groundwater input of CO 2 -rich waters (McDonald et al., 2013).Lake metabolism also impacts δ 13 C DIC dynamics.Previous studies have shown that δ 13 C of DIC in lakes is driven by carbonate chemistry, hydrology (i.e., groundwater inflow), and metabolic activity (Bade et al., 2004).Primary production increases δ 13 C DIC because of the preferential uptake of 12 C (isotope fractionation), while organic matter respiration decreases δ 13 C DIC (Fry, 2006).
If pCO 2 and δ 13 C DIC would have been only or primarily controlled by the balance between respiration and production of organic matter, one would expect a tight correlation between δ 13 C DIC and pCO 2 , which was not observed overall (Fig. 3A, Table 2), indicating that other factors are involved.High inorganic carbon loadings of inflowing rivers and groundwater inputs may sustain the CO 2 excess (McDonald et al., 2013) and govern the δ 13 C DIC (Bade et al., 2004).
Moreover, carbon dioxide water-air exchange reactions may have modified δ 13 C DIC values because of isotope fractionation during water-air exchange, in particular at high pH/low CO 2 values (Herczeg and Fairbanks, 1987;Bade et al., 2004;Bontes et al., 2006).
The correlation between δ 13 C DIC and pCO 2 in oligo-mesotrophic lakes was stronger (Table 2) and pCO 2 was highest and δ 13 C DIC was most depleted in 13 C at the highest DOC in oligo-mesotrophic lakes (Fig. 1C, 3C).Such a depletion of δ 13 C DIC with increasing DOC, as an indicator of the importance of respiration in oligo-mesotrophic lakes, has been shown previously in North-American lakes (Lennon et al., 2006).In addition to community respiration, methanotrophic bacteria in high DOC lakes could decrease δ 13 C DIC (Jones et al. 1999).However, anoxic hypolimnia are rare in these lakes, either due to low nutrients or polymixis, indicating that methanogenesis was not of major importance in the lakes investigated.Furthermore, we examined the δ 13 C of fatty acids abundant in or specific to methanotrophs and these were not more depleted in 13 C than other fatty acids.
Note that the lakes were only sampled at one point location and depth, representing average conditions, so spatial variability per lake is not taken into account.Also diurnal variation and variation over the year in each lake are not considered in this study.However, the aim of our study is comparing snapshots of different lakes representing different metabolic states and not to describe the biogeochemistry of each individual lake.So, although we miss some variation, this shouldn't affect our main findings on carbon isotope biogeochemistry of these lakes.

Allochthonous δ 13 C
Dissolved and particulate organic carbon pools are mixtures of organic matter from various sources with potentially different stable carbon isotopic compositions.The particulate organic carbon pool comprises biomass from living organisms and remains from organisms within the lake as well as allochthonous detritus.The relative importance of living biomass to total POC pool, calculated based on equation 4, varies from 5.7 to 93 % (Table 1), with on average about 53± 20 % in eutrophic lakes and only 13±5 % in in oligo-mesotrophic lakes.In our study we have explicit carbon isotope data for the most important living compartments (algae, bacteria and zooplankton; De Kluijver, 2012), but we have no direct measurement of the δ 13 C of organic carbon delivered from the watershed via atmospheric, riverine and groundwater inputs.We have therefore used two proxies for δ 13 C allo ; the carbon isotope ratio of dissolved organic carbon and that of detrital particulate organic carbon calculated by difference (equations 3, 4).Both proxies for δ 13 C allo provide an estimate for the total detrital pool, i.e. the sum of aquatic and terrestrial detritus.
The oligo-mesotrophic lakes are surrounded by forest (C 3 vegetation) and δ 13 C DOC was -28.0 ± 0.5 ‰, corresponding to a C 3 vegetation signal, suggesting that the dissolved organic carbon pool is dominantly terrestrially derived, consistent with a combined carbon and hydrogen isotope study of Wisconsin and Michigan lakes by Wilkinson et al., (2013).The other proxy for allochthonous carbon, δ 13 C det, was slightly more negative (-29.6 ± 2.1 ‰), but the two proxies for allochthonous carbon were well correlated (Table 2).The 1.6 ‰ lighter isotopic composition might reflect a relatively larger contribution of autochthonous detritus to the total detrital particulate organic matter pool than to the dissolved pool.Consistently, Wilkinson et al., (2013) reported that a lower contribution of terrestrial organic matter to the particulate pool than to the dissolved organic matter pools in North American lakes.
Allochthonous carbon proxies in eutrophic lakes were more 13 C-enriched and variable: -25.4 ± 1.1 ‰ for δ 13 C DOC and -26.6 ± 4.2 ‰ for δ 13 C det .Moreover, δ 13 C DOC and δ 13 C det were not significantly correlated.The enrichment in 13 C of δ 13 C allo in eutrophic lakes can be partly explained by land use in the water shed; almost all eutrophic lakes were located in the state of Iowa, where an average of 92% of the land is under periodic cultivation for maize (C 4 plants, -14 ‰).There was more uncertainty in δ 13 C allo in the eutrophic lakes for two main reasons.First, we expect a substantial autochthonous contribution to DOC and detritus in productive lakes (Bade et al., 2007), which contributes to the larger range in δ 13 C DOC and δ 13 C det (Fig. 2).Second, the presence of C 3 and C 4 vegetation with their distinct isotopic compositions can create a variable δ 13 C allo .Variability in δ 13 C allo has received far less attention than that of aquatic primary producers.Our results show distinct differences in the isotopic composition of external subsidies and argue against a fixed value for allochthonous carbon, especially in areas with abundant C 4 vegetation, such as maize.

Phytoplankton δ 13 C
The determination of δ 13 C phyto is one of the major challenges in aquatic ecology.Fatty acid biomarkers as proxies for δ 13 C phyto have the advantage that there is a larger certainty that measured δ 13 C values represent parts of phytoplankton carbon.The main uncertainty using δ 13 C FA as marker for δ 13 C phyto comes from the isotopic offset between lipids and total cells (Δδ 13 C FA-cell) which depends on species composition (summarized in, e.g., Hayes, 2001), growth conditions (e.g., Riebesell et al., 2000) and the FA considered (Fig. 5).
Isotope fractionation between CO 2 and phytoplankton was variable (8 -25 ‰) in our study (Table 3).This implies that calculations of δ 13 C phyto from 2 CO 13 C  with a constant fractionation factor provides inaccurate results, consistent with methodological comparisons by Marty andPlanas, (2008) andMcCallister et al., (2008).The usual value for photosynthetic fractionation in phytoplankton is ~20 ‰, based on C 3 photosynthesis (Fry, 2006), but several studies that determined ε in lakes showed that actual fractionation is usually lower than this value (Cole et al., 2002;Bade et al., 2006).Also, in our study, fractionation was lower (~17 ‰) on average, and highly variable, especially in eutrophic lakes.There are several explanations for this variability.1) Actual fractionation has been shown to be dependent on several variables, including growth rate (Bidigare et al., 1997) and CO 2 availability (Laws et al., 1995).
Fractionation is highest under high CO 2 availability and low growth rates.In the less productive oligo-mesotrophic lakes, the conditions favor optimal fractionation, and therefore, fractionation was rather constant (Fig. 5C).In the productive, eutrophic lakes, actual fractionation was influenced by pCO 2 and C phyto , with lowest fractionation and therefore most enriched 13 C phytoplankton in the most productive (low CO 2 and high C phyto ) lakes (Fig. 5A, 5B).
Two clusters in δ 13 C phyto were present in the eutrophic lakes (Fig. 5A, 5B) and the shift occurred when lakes were below 20 µmol L -1 CO 2 in the eutrophic lakes.When CO 2 becomes limiting, some phytoplankton can also shift to bicarbonate utilisation, which is isotopically enriched by ~8 ‰ compared to CO 2 .Direct uptake of carbonate and conversion in the carboxysomes is very common in the Cyanobacteria that dominate eutrophic lakes (Bontes et al., 2006).The lakes with 13 C-enriched phytoplankton had high concentrations of zeaxanthin, a biomarker for cyanobacteria.However, the higher δ 13 C phyto in high zeaxanthin lakes was not a direct consequence of 13 C-enrichment in Cyanobacteria.FA that are abundant in Cyanobacteria (18:nωn) were not more enriched than FA that are absent in Cyanobacteria; in fact, they were the most 13 C-depleted of all FA (Fig 4).Cyanobacteria grown in laboratory cultures also showed higher fractionation (up to 9 ‰) in lipids relative to total biomass than eukaryotic phytoplankton (summarized in Hayes, 2001).
The most 13 C-enriched phytoplankton FA was 22:6ω3, which is abundant in dinoflagellates (Fig. 4) (Dalsgaard et al., 2003).Dinoflagellates were also more enriched in 13 C compared to other phytoplankton in a subtropical lake (Zohary et al., 1994).A possible explanation for 13 C-enriched dinoflagellates in field studies, can be their mixotrophic character, so that part of their isotopic composition reflects consumer δ 13 C.However, PUFAs of autotrophic dinoflagellates grown in continuous cultures were also more 13 C-enriched to C16:0 than PUFA of other phytoplankton (Schouten et al., 1998).
Finally, variability in Δδ 13 C FA-cell can contribute to the observed variability.In laboratory studies, the offset between lipids and bulk material has shown to be variable (van Dongen et al., 2002, Fiorini et al., 2010).One can expect that in field studies, with multiple species, however, these cellular variations would probably disappear within broader trends.If we assume an overall mean Δδ 13 C FA-cell , then the uncertainty in the actual value would affect the absolute fractionation values, but not the observed variability in fractionation.

Carbohydrates and lipid δ 13 C
The enrichment in 13 C of carbohydrates and depletion in 13 C of lipids relative to total cells (mainly amino acids) has been shown in culture studies of phytoplankton (Van Dongen et al., 2002) and in culture studies of several primary producers and consumers (Teece and Fogel, 2007).Results of this study show that the enrichment in 13 C in glucose as well as the 13 Cdepletion in in fatty acids relative to bulk material can also be detected in field samples (Fig. 2, Table 3).We observed that tot FA -gluc 13 C   increased with TP (Table 2), but whether this represents a general phenomenon for lakes needs further exploration.
Bacterial FA were more enriched in 13 C than phytoplankton FA in all lakes (Fig. 2, Fig. 4).This observation can be explained by differences in carbon source or differences in Δδ 13 C FA- cell between phytoplankton and bacteria.Carbohydrates, present in high concentrations in DOC, form an important carbon source for bacteria.Since carbohydrates were the most 13 C-enriched carbon source, a preferential use of carbohydrates, would result in 13 C-enriched bacteria (Fig. 2).
Another explanation is that isotope fractionation during FA synthesis was smaller in bacteria compared to phytoplankton.There are no field studies on Δδ 13 C FA-cell in freshwater bacteria, but field studies on sediment and marine bacteria report a range of 0 -5 ‰ in Δδ 13 C FA-cell (Hayes, 2001, Burke et al., 2003;Bouillon and Boschker, 2006).Burke et al., (2003) suggested that in field samples with complex communities and substrates, Δδ 13 C FA-cell would be ~0 ‰.The results of our study support this idea, since bacterial FA had a similar δ 13 C as POC.If a similar Δδ 13 C FA- cell for phytoplankton and bacteria would be used, bacteria would be more enriched in 13 C than its potential carbon sources in half of the studied lakes, which is rather unlikely.

Conclusions
Our results show that trophic state has a large influence on lake metabolism and carbon cycling in plankton food webs.Overall, eutrophic lakes had larger variability in δ 13 C in all organic carbon pools than oligo-mesotrophic lakes, caused by larger isotopic variability in the base of the food web in eutrophic lakes (both allochthonous and autochthonous carbon).In eutrophic lakes, δ 13 C phyto showed that two clusters of phytoplankton were present, with the most 13 C-enriched phytoplankton at high CO 2 and high chl a. Dominance of cyanobacteria played a role, but enrichment in 13 C was present in all phytoplankton, as seen in specific PLFA.

Figure 2 :
Figure 2: Box and whisker plot of the δ 13 C of inorganic and organic carbon pools in eutrophic lakes (grey boxes, n=11) and oligo-mesotrophic lakes (white boxes, n=11).The dashed lines present typical values for C 3 and C 4 vegetation.

Figure 5 :
Figure 5: The relation of δ 13 C phyto in eutrophic lakes (filled circles, n=11) and oligo-mesotrophic lakes (open circles, n=11) to A) C phyto and B) pCO 2 .Panel C presents a box whisker plot of calculated

Table 1 :
Limnological characteristics of the sampled lakes.pCO 2 was determined from temperature, DIC, and pH.C phyto presents the average of chl a and fatty acid based phytoplankton biomass.C bac presents fatty acid derived bacteria carbon biomass.

Table 3 :
Carbon isotope values of sampled lakes.Isotope data are presented as average ± sd (n=3).DIC (equation 1).δ 13 C phyto and δ 13 C bac are not corrected for the offset between fatty acids and total cells, but