Introduction

Decomposition of organic matter makes available the products of primary photosynthesis to detritus-based food webs and releases inorganic components into the environment. Multiple studies have examined detrital mass-loss rates, showing generally positive relationships with nutrient content and negative relationships with concentrations of recalcitrant compounds in the litter, relationships that tend to hold across many ecosystems (Melillo et al. 1984; Taylor et al. 1989; Enriquez et al. 1993; Parton et al. 2007; Cornwell et al. 2008).

In stream and rivers, decomposition rates are rapid due to constant water availability and nutrient supply (Tranvik et al. 2009; Battin et al. 2009). Leaf litter inputs are essential resources for forested headwater streams (Vannote et al. 1980), which strongly affect food webs (Wallace et al. 1997; Gessner et al. 2010) and ecosystem function (Fisher and Likens 1973; Mulholland et al. 1985, 2000; Meyer et al. 1998; Argerich et al. 2008). Increasing stream nutrient concentrations accelerate detrital mass loss (Meyer and Johnson 1983; Suberkropp and Chauvet 1995; Gulis and Suberkropp 2003; Ferreira et al. 2014), although this effect can be reversed at high nutrient concentrations (Carreiro et al. 2000; Woodward et al. 2012). Variation in relationships among decomposition rates, leaf characteristics (litter quality), and stream nutrient concentrations has been partially explained by different responses in biomass accrual or activity of microbial assemblages (hereafter referred to as ‘biofilms’) on leaf litter (Gessner and Chauvet 1994; Gessner 1997; Gulis and Suberkropp 2003; Stelzer et al. 2003).

Litter decomposition in streams is usually measured as net loss of litter mass and net changes in its element content over time (Tank et al. 2010). However, changes in element content are the result of simultaneous gross fluxes of elements released from and retained in the litter. Processes driving litter element loss include chemical leaching, microbial mineralization of organic matter, physical fragmentation, and breakdown by stream animals. Additionally, biofilms growing on litter assimilate organic C and N from the leaf and from streamwater, which can be further lost from the biofilm-litter system through respiration and deamination (i.e., N mineralization). Concurrently, C and N gross fluxes from streamwater into the biofilm-litter system take place due to silt deposition and abiotic adsorption (Bott et al. 1984; Webster and Benfield 1986).

Despite the common empirical observations of the positive effects of stream nutrient concentrations on decomposition rates (Ferreira et al. 2014), studies quantifying gross fluxes entering the biofilm-litter are scarce. Moreover, most of the available data on immobilization is for bulk detrital samples and for a single point in time (e.g., Tank et al. 2000; Dodds et al. 2004). How immobilization varies throughout decomposition and in response to different leaf litter types is not known, even though the chemical properties of leaf litter vary strongly among species and over time (Cornwell et al. 2008). Finally, most studies focus on one element (i.e., N), even though biosynthesis couples N with C in specific ratios with effects scaling to the ecosystem (Sterner and Elser 2002; Gruber and Galloway 2008). In this study, we aim to fill some of these gaps in our knowledge by evaluating the immobilization fluxes into the biofilm-litter during decomposition of two closely related tree species, which differ strongly in phytochemical characteristics.

Researchers have provided several lines of evidence to explain biofilm immobilization of dissolved organic C (DOC) and dissolved inorganic and organic N from streamwater. First, C and N stoichiometry of litter frequently does not fulfill the elemental requirements of biofilms because microorganisms have lower C:N ratios than the litter substrate (Sterner and Elser 2002; Parton et al. 2007). Second, recalcitrant compounds in litter decrease microbial enzymatic activity (Sinsabaugh et al. 1993) and the proportion of less readily resources available for biofilms (Gessner and Chauvet 1994). Thus, shortage of suitable resources might enhance biofilm DOC and N uptake from the water column. Third, DOC and N in the water could occur in readily available forms and are thus easily assimilated (Wiegner et al. 2005; Kaplan et al. 2008). Therefore, differences in litter characteristics may influence nutrient immobilization from streamwater during decomposition. Finally, biofilms may vary in microbial biomass and composition depending on litter types (Wymore et al. 2013; Frossard et al. 2013). The accumulation of microbial biofilm on the decomposing leaf increases the capacity to assimilate elements from the surrounding environment by presenting a higher assimilating surface area to the water column. Furthermore, the nature of this assimilating surface could influence the stoichiometry of microbial biosynthesis and the need to import elements from streamwater into the leaf-biofilm complex.

In this study, we quantified the relative importance of the C and N fluxes from streamwater into biofilm on litter. We used 13C- and 15N-enriched leaf litter and applied a variation of the isotope pool dilution method (Kirkham and Bartholomew 1954), which has been widely used in soil biogeochemistry to study nutrient dynamics during decomposition in soils (Murphy et al. 2003). This method quantifies the rate at which the isotopic value of an artificially enriched element pool declines due to the mass fluxes from an unlabeled pool (Kirkham and Bartholomew 1954).

We used 13C- and 15N-labeled litter from two foundation riparian tree species, Populus fremontii and Populus angustifolia. Phytochemical differences (Table 2) between these species, especially tannin content, drive changes in their decomposition rates with implications for adjacent terrestrial and aquatic ecosystems (Driebe and Whitham 2000; LeRoy et al. 2006; Whitham et al. 2006; Holeski et al. 2012). We expected that microbes growing on litter with a higher content of recalcitrant compounds would show a relatively greater reliance on C and N from streamwater than those growing on leaves with a lower content of recalcitrant compounds because these compounds are a less-accessible resource for heterotrophic microbes. Understanding the relative importance of C and N sources for biofilms on litter and how they vary during decomposition and among litter types will yield insights on the mechanisms driving litter decomposition, how decomposition controls the flux of C and N to the microbial food web, and the basic microbial and chemical controls on stream biogeochemical cycling.

Materials and methods

Study site

This study was conducted in upper Oak Creek (1600 m a.s.l) on the southern edge of the Colorado Plateau (35°02′N, 111°43′W; Arizona). Oak Creek is a first-order stream in a deep narrow canyon. It drains a 77,450-km2 catchment extensively covered by ponderosa pine (Pinus ponderosa). This area is characterized by steep topography and limestone and sandstone bedrock (LeRoy et al. 2006; Wymore et al. 2013). The riparian vegetation is predominately deciduous, including Fremont cottonwood (Populus fremontii), narrowleaf cottonwood (Populus angustifolia), Arizona alder (Alnus oblongifolia Torr.), Arizona sycamore (Platanus wrightii S. Wats.), coyote willow (Salix exigua Nutt.), and Goodding’s willow (Salix gooddingii Ball) (LeRoy et al. 2006). This experiment was conducted from November to December 2011. During this time, discharge, streamwater temperature, pH, O2 concentration and specific conductivity were relatively constant and concentrations of stream nutrients and isotopic values of dissolved N and DOC were low (Table 1).

Table 1 Physical and chemical parameters measured at Oak Creek during the experimental period (range of values or SE in parentheses)
Full size table

Field experiment with labeled leaf litter

Tree cuttings of P. fremontii and P. angustifolia, from the Ogden Nature Center common garden (Ogden, UT) were grown at the Northern Arizona University Arboretum Research Greenhouse. Plants were isotopically labeled in a hydroponic nutrient solution by dripping (15NH4)2SO4 in aqueous solution into pots twice a week and pulsed with 99 atom % 13CO2 for 4 h twice a week for 6 months (Compson et al. 2014, in revision). Variability in litter chemistry is lower for P. fremontii compared to P. angustifolia (Wymore et al. 2013; Table 2), so we selected five genotypes of P. fremontii and ten of P. angustifolia. Naturally senesced leaf litter was collected, air dried and stored. The variability of the stable C isotope composition (δ13C) and stable N isotope composition (δ15N) label in different leaf litter tissues was tested by boiling individual pieces of leaf litter (n = 10 per species) for 1 h to remove nearly all the soluble compounds and then taking the difference between pre- and post-leached samples. No differences between litter species (P. fremontii and P. angustifolia) for δ13C (t-test t 1,18 = 0.16, P = 0.87) or δ15N (t-test t 1,18 = −0.16, P = 0.87) were found (Compson et al. 2014, in revision).

Table 2 Initial litter characteristics and decomposition dynamics for Populus fremontii and Populus angustifolia (mean and SE)
Full size table

For each genotype, litter was mixed and three composite samples were analyzed for initial C and N content and isotope composition using a Carlo Erba NC 2,100 Elemental Analyzer (CE Instruments, Milan) interfaced with a Thermo-Finnigan Delta Plus XL (Thermo-Electron, Bremen, Germany) isotope ratio mass spectrometer (IRMS) at the Colorado Plateau Stable Isotope Laboratory (CPSIL) (http://www.isotope.nau.edu). P. angustifolia litter (high-tannin litter), had higher % C values than P. fremontii litter (low-tannin litter), but % N and C:N did not statistically differ (Table 2).

Litter was incubated in the stream using fine-mesh litterbags (10.5 × 10.5 cm2, 0.5-mm mesh), which were deployed in the river zip-tied to rebar on 10 November 2011. Each litterbag contained 1 g of leaf litter. After 6, 13, 20 and 27 days of the experiment, 15 litterbags of P. fremontii and 30 litterbags of P. angustifolia (3 replicates × genotype) were collected from the stream (only P. angustifolia litterbags were collected for the final harvest). Upon harvest, litterbags were placed into zip-lock bags, and transported on ice to the laboratory where they were processed within 24 h.

For each harvest, dissolved O2 (DO), conductivity, pH and water temperature were determined using a Hydrolab Minisonde (Hydrolab-Hach, Loveland, CO) in a five-point transect along the experimental reach. Stream discharge data were obtained from the United States Geological Survey Oak Creek weather station. Three replicate water samples (~4 L each) were collected upstream from the experimental reach on day 13 of litterbag incubation and analyzed for NO3 − and DOC concentration and isotope composition. Water was filtered through a 0.2-µm Acrodisk filter and analyzed colorimetrically for NH4 + and NO3 − concentration using an autoanalyzer (Lachat Quickchem FIA+ 8000; Lachat Instruments, Milwaukee, WI). DOC concentration was analyzed by the persulfate oxidation method with an OI Analytical Model 1,010 Total Carbon Analyzer connected to a Delta Plus Advantage IRMS. The δ15N of NO3 − was determined by reduction to N2O followed by coupled gas chromatography (Thermo Finnigan Precon and Delta Advantage IRMS), using the bacteria denitrification method (Casciotti et al. 2002) at CPSIL. The δ15N of NH4 + was determined using the ammonia-diffusion method (Holmes et al. 1998).

Laboratory analysis

Litter was removed from the litterbags, rinsed with deionized water, and wet mass was recorded. At each harvest date, litter content from the three replicate bags was pooled and homogenized, resulting in five composite samples for P. fremontii and ten for P. angustifolia. Each composite sample was subsequently split into two subsamples, one for bulk-litter elemental analysis (~1 g wet weight), and the rest for determination of microbial biomass.

Percent moisture of litter was determined for bulk litter by weighing subsamples before and after oven drying at 60 °C for 24 h. Dried litter was ground with a mortar and pestle to a fine powder and was analyzed for C and N content and isotopic composition at CPSIL, as described above. Subsamples for microbial biomass determination were processed using an adaptation of the chloroform fumigation-extraction technique, originally developed for soils (Brookes et al. 1985; Vance et al. 1987) and later modified for stream detritus (Mulholland et al. 2000; Sanzone et al. 2001; Cheever et al. 2013). Labile organic material was extracted from the litter with 50 mL of 0.05 M K2SO4, stored on ice overnight, shaken for 1 h, and centrifuged at 9,800 g for 10 min, after which the supernatant was poured off and discarded. Litter samples were then placed in glass beakers in a desiccator and fumigated with alcohol-free chloroform. The desiccator was evacuated until chloroform boiled. Samples were vented three times, and then sealed under vacuum and kept in the dark for 24 h. Fumigated samples were then removed from the desiccator, extracted with 50 mL of 0.05 M K2SO4, shaken for 1 h, and centrifuged at 9,800 g for 10 min. The supernatant was filtered through 1.2-µm filters (Supor Membrane; PALL Live Sciences, NY) and placed in a ventilated oven (60 °C) for 48 h. Dried K2SO4 salts from both pre- and post-fumigation samples were ground with a mortar and pestle to a fine powder, weighed, and analyzed for C and N elemental and isotope composition as described above. K2SO4 blanks routinely yield non-detectable C and N peaks using this method. To calculate immobilization rates, 15N and 13C isotopic values were expressed in atom % excess, correcting for the natural abundance of C and N isotopes, respectively; that is, 13C atom (atom % excess) = 100 × [13C/(13C + 12C)]−1.1112 % and 15N atom (atom % excess) = 100 × [15N/(15N + 14N)]−0.3663 %. For salt samples, the precision of the international standard NIST 2711 MT soil was ±5.0 × 10−7 atom % excess for 13C and ± 6.9 × 10−7 atom % excess for 15N (SD of six replicate samples). The precision of the NIST peach leaves standard, when bulk litter samples were run, was ± 6.1 × 10−6 atom % excess for 13C and ± 2.5 × 10−4 atom % excess for 15N (SD of 32 replicate samples).

Parameter calculations

To characterize decomposition rates for litter types, we calculated the leaf litter decomposition rate constant (k, day−1) as the slope of the log-transformed percentage of remaining mass over time for each Populus genotype (Benfield 2006). Microbial biomass, in terms of C (MBC) and N (MBN), was estimated using the C and N content in the fumigated litter samples for each date. MBC and MBN were expressed per unit of litter mass (i.e., mg C or N g−1 litter) to compare results among harvest dates and between the two litter types. The change in isotopic and N isotope composition in the fumigated litter samples was used in a two-end member mixing model to quantify the relative contribution of leaf litter and water column C and N to microbial biomass (Phillips and Gregg 2003; Boecklen et al. 2011). For each harvest date, the percentage of C or N in microbial biomass derived from streamwater was calculated using the following equations:

$$\%\,C {\text{streamwater source}} = \frac{{^{13} C\, {\text{microorganisms}} - ^{13} C\,{\text{leaf}} }}{{ ^{13} C\,{\text{DOC}} - ^{13} C\, {\text{leaf}}}} \times 100$$
(1)
$$\%\, N {\text{streamwater source}} = \frac{{^{15} N\, {\text{microorganisms}} - ^{15} N\, {\text{leaf}}}}{{^{15} {N}\, {\text{NO}}_{3}^{ - } -^{15} N\, {\text{leaf}}}} \times 100$$
(2)

where 13 C and 15 N in microorganisms are the isotopic values (in atom  %) of the chloroform-extracted fraction, 13 C and 15 N in leaves are the initial isotope values (in atom  %) measured in the leaves, and 13C DOC and 15N NO3 − are the isotope values (in atom  %) measured in the streamwater samples (Table 1). Thus, the percentages of C and N in microbial biomass derived from leaf sources are the remaining percentages, i.e., % C leaf source = 100 % C streamwater source and % N leaf source = 100 % N streamwater source. Equation 2 assumes that the isotopic values of NO3 − are similar to those of other streamwater N sources in relation to the high isotopic label of the leaf litter. This assumption is in agreement with the results of Eq. 2 using NH4 +-15N instead of NO3 −-15N, which did not significantly alter the results.

To quantify C and N fluxes from the water column into the biofilm-litter system on each harvest date, gross immobilization (GI) rates of C (GIC; mg C g−1 litter day−1 ) and N (GIN; mg N g−1 litter day−1) were calculated using the isotope pool dilution method (Kirkham and Bartholomew 1954; Hart et al. 1994):

$$GI_{{(\Delta t)}} = \frac{{M_{i} - M_{f} }}{{t_{f} - t_{i} }} \times \frac{{\log _{{10}} ~\frac{{M_{{i~at\% ~excess}} }}{{M_{{f~at\% ~excess}} }}}}{{\log _{{10}} ~\frac{{M_{i} }}{{M_{f} }}}}$$
(3)

where M is the mass of C or N in the biofilm-litter system at the initial (t i) and final (t f) time (in mg C or N g−1 litter) and M i at% excess and M f at% excess are the 15N or 13C atom % excess of the biofilm-litter system for the same interval. Therefore, the measurement of GI is based on the dilution of the isotope composition of the biofilm-litter system over time as a reflection of the import of C and N from the unlabeled water column pools. In order to integrate the isotopic data from all harvests, we generated linear models of decay of 15N and 13C isotope for each replicate, and then used the measured isotopic values of the litter (atom % excess) for the initial and final pool, for the interval of time studied, as input into Eq. 3. Finally, GI rates of C and N were standardized by MBC and MBN, respectively, as a measurement of microbial efficiency for GI of these elements, which can be compared between litter types and among sampling dates.

Data analysis

Parameters were calculated for each cottonwood genotype; each genotype at each sampling date was treated as a replicate for each species. Two-group Student’s t-tests were used to compare values of k between litter types. Analyses of covariance (ANCOVAs) were used to analyze the effect of litter type, with harvest days as an additive covariate, on MBC and MBN, % of microbial mass derived from the leaf source for both C and N, and GIC and GIN. To test for differences between litter types in the GIC:GIN, we bootstrapped the difference of these ratios between litter types (1,000 iterations for each time point and species), determined the 95 % confidence intervals (CIs) for this difference, and determined whether it overlapped with zero. The correlation between the ratio GIC:GIN and microbial C:N was tested using Spearman correlation. The Resample Stat add-in for Excel software was used for the bootstrapping procedure (http://www.resample.com/excel/). All other statistical analyses were conducted using R, version 2.15.1 (R Development Core Team 2012).

Results

Decomposition rates and microbial biomass

As expected, leaf litter decomposition rate was 1.5 times higher for low-tannin litter (P. fremontii) than high-tannin leaf litter (P. angustifolia; Table 2). For the two litter types, microbial biomass, either measured as C or N biomass, increased until day 13, but then leveled off (Fig. 1a, b). Microbial C:N showed the opposite pattern, decreasing sharply at the second harvest, then leveled off. Microbial biomass (both as MBC and MBN) per gram of litter was higher in litter with a low tannin content than in litter with a high tannin content (ANCOVA, MBC F 1,42 = 32.09, P < 0.0001; MBN F 1,42 = 38.38, P < 0.0001; Fig. 1a, b). In contrast, microbial C:N was higher in P. angustifolia litter than in P. fremontii litter (ANCOVA, F 1,42 = 12.39, P < 0.01; Fig. 1c). Microbial C accounted for 2.7–8.4 % of the total litter C pool in P. fremontii litter and between 1.1 and 5.3 % for P. angustifolia litter. Microbial N represented between 4.7 and 18.5 % for P. fremontii litter and between 2.1 and 15.0 % for P. angustifolia litter.

Fig. 1
figure 1

Temporal variation of microbial biomass (MB) C (MB C ; a), N (MB N ; b) and C:N mass ratio (c) during the leaf litter decomposition period for Populus fremontii (grey circles; n = 5) and Populus angustifolia (black circles; n = 10). Data points are means and vertical bars represent SEs

Full size image

Contribution of microbial C and N from the streamwater

The proportion of C and N derived from streamwater increased during the incubation, and at the last harvest (27 days) accounted for 32 % for C and 38 % for N (average values for P. angustifolia) of the microbial biomass (Fig. 2). For both litter types, and averaged over the duration of the experiment, leaf litter was the major source of C and N for the growth of microbial assemblages (average over the two litter types: 89 ± 2 % for C, 81 ± 3 % for N). We did not find significant differences between leaf types in the percentages of C and N in microbial mass that were derived from the streamwater (ANCOVA: P > 0.05).

Fig. 2
figure 2

Temporal variation of the percentage of C (a) and N (b) in the microbial assemblage that is derived from the streamwater for P. fremontii (grey circles; n = 5) and P. angustifolia (black circles; n = 10). Due to method error, some percentages are lower than 0 %. Data points are means and vertical bars represent SEs

Full size image

GIC and GIN from streamwater into the biofilm-litter system

GIC was, on average, almost two times higher for P. angustifolia litter (GIC = 3.79 ± 0.41 mg C g−1 litter day−1) than for P. fremontii litter (GIC = 1.94 ± 0.59 mg N g−1 litter day−1; Fig. 3a; ANCOVA, F 1,42 = 5.55, P < 0.05). The pattern reversed for GIN, which was, on average, two times higher for P. fremontii litter (GIN = 0.08 ± 0.02 mg N g−1 litter day−1) compared to P. angustifolia litter (GIN = 0.16 ± 0.02 mg C g−1 litter day−1; Fig. 3b; ANCOVA, F 1,42 = 6.82, P < 0.05). The ratio between GIC and GIN was significantly higher for P. angustifolia litter (on average 35.2) than for P. fremontii litter (on average 13.9) (95 % CI for the difference, 6.5–59.6; Fig. 3c). GIC standardized by the microbial C content was nearly three times higher for P. angustifolia litter (mean 0.40 ± 0.06 mg C mg−1 MBC day−1) than for P. fremontii litter (mean 0.12 ± 0.05 mg C mg−1 MBC day−1; ANCOVA, F 1,42 = 7.09, P < 0.05). In contrast, there was no difference between litter types for GIN standardized by the microbial N content (average for both species 0.07 ± 0.01 mg N mg−1 MBN day−1; ANCOVA, P > 0.05). Finally, microbial GIC:GIN was not related to microbial C:N for either litter species (Spearman correlation; P > 0.05).

Fig. 3
figure 3

Temporal variation of gross immobilization (GI) rates for C (GI C ; a) N (GI N; b) and their stoichiometric relationship (c) for P. fremontii (grey circles; n = 5) and P. angustifolia (black circles; n = 10). Data points are means and vertical bars represent SEs

Full size image

Discussion

The main goal of this study was to understand the biogeochemical interactions between the biofilm litter system and the streamwater during litter decomposition. Our results indicated that litter phytochemical characteristics had a strong effect on the biomass and stoichiometry of microorganisms growing on litter. The strong species differences in C and N immobilization from the water column into the litter biofilm (Fig. 3) suggest that microbial assemblages on these species differ in C and N demand, and that these differences are strongly influenced by leaf litter characteristics.

Litter decomposition and microbial biomass

Decomposition rates were higher for P. fremontii than P. angustifolia, as previously shown (Driebe and Whitham 2000; LeRoy et al. 2006; Holeski et al. 2012). Leaf litter with a lower content of recalcitrant compounds (P. fremontii) accrued more microbial biomass compared to litter with a higher content of recalcitrant compound (P. angustifolia). This finding is in agreement with results from previous studies where recalcitrant litter types showed low microbial biomass accrual (Gulis and Suberkropp 2003; Talbot and Treseder 2012; but see LeRoy et al. 2007). In addition, elemental stoichiometry of biofilms differed between litter types, with higher C:N values for high-tannin litter. Differences in C:N ratio among microbial biofilms on litter might be explained by differences in the composition of the microbial assemblage, as fungal mycelia often have higher C:N ratios than bacteria (Sterner and Elser 2002; Strickland and Rousk 2010). Leaf-associated fungal biomass typically exceeds bacterial biomass (Gessner 1997; Findlay et al. 2002), and our results suggest that this pattern is stronger for biofilms on high-tannin compared to biofilms on low-tannin litter. This fact is further supported by a related study on decomposing cottonwood litter in the same stream reach where quantitative polymerase chain reaction results revealed a higher fungi:bacteria gene abundance ratio for P. angustifolia than for P. fremontii (Wymore et al. 2013). Fungi may be better competitors than bacteria in more recalcitrant leaves due to their hyphal networks and enzymatic capabilities to break down recalcitrant materials compared to bacteria (Kohlmeier et al. 2005; De Boer et al. 2005; Moorhead and Sinsabaugh 2006; Romaní et al. 2006).

The relative contribution of C and N from streamwater during decomposition

The reliance on elemental resources from streamwater by biofilms was low at the beginning of the decomposition process and increased with time for both litter types, probably as labile compounds were consumed by microbes or leached out of the leaves. The fact that immobilization fluxes and microbial biomass remained fairly constant over time would explain the linear increase of the contribution of C and N from the streamwater. Previous studies reported similar patterns. For example, Cheever et al. (2013) showed that microorganisms colonizing decomposing leaves acquired more N from the streamwater during late decomposition stages compared to early stages, deriving up to 80–90 % of N from the water column by the end of the decomposition experiment (i.e., 12–15 weeks). In other systems, such as a N-rich estuary, microbial assimilation of DIN into particulate organic material also increased with time, reaching nearly 70 % by the end of the experiment (Caraco et al. 1998). Information about microbial reliance on streamwater for C is mostly limited to microorganisms in sediments, where DOC has been estimated to support up to half of their metabolism (Findlay et al. 1993; Fischer et al. 2002; Sobczak and Findlay 2002; Wiegner et al. 2005). Considering C and N together, our data suggest that N derived from streamwater is an important supplement for microbial growth, supporting the common assumption that stoichiometric constraints faced by microorganisms growing on litter would require N from streamwater (Sterner and Elser 2002). Our data also indicate that streamwater organic C is an important source for microbial communities, almost as important as N. The importance of stream C supplements during leaf litter decomposition is surprising (Cheever et al. 2013) and suggests high microbial reliance on multiple streamwater resources during decomposition.

Immobilization of C and N into the biofilm-litter system: contrasting patterns between litter types

Leaf species differed in the stoichiometry of C and N fluxes from the streamwater to the biofilm-litter system. Gross immobilization rates of N were higher on low-tannin litter compared to litter with high-tannin content, contrary to our expectations, probably because a higher content of recalcitrant compounds in the latter slowed down microbial growth and consequently reduced N demand from the streamwater. This argument is supported by specific rates of N immobilization (per unit microbial biomass) which did not differ between litter types.

In contrast, gross immobilization rates of C were higher in high-tannin litter, even when standardized by microbial biomass, indicating higher import of C from streamwater into the biofilm-litter compartment in high-tannin litter. Because tannin compounds may form relatively recalcitrant complexes, it is reasonable to think that C in the high-tannin litter is a less accessible resource, such that microorganisms obtain C more efficiently from the water column. Thus, observations for the C immobilization rate support our hypothesis that the concentration of recalcitrant compounds in litter would increase the dependence on streamwater by microbial biofilms. Furthermore, the decoupling between immobilization C:N from streamwater and microbial C:N further suggested that the interaction of biofilm-litter with streamwater is dependent on contrasting C and N contributions of the leaf litter resource.

Despite the fact that heterogeneous isotopic label in the leaf litter could have increased uncertainty of our immobilization estimations, differences between species were higher, indicating the robustness of the patterns found. Moreover, our estimations might have also included other inputs of C and N besides active uptake by microbes, such as abiotic chemical adsorption and deposition. Despite these pitfalls, the isotopic dilution observed over time suggested the relevance of the biotic uptake over these other processes. Because algae might also have an active role in litter decomposition (Danger et al. 2013; Rier et al. 2014), photosynthesis activity would have resulted in overestimations of microbial immobilization. Even so, severe shading due to reach geomorphology likely limited algae activity (J. Marks, personal observation), diminishing the influence of algae on our estimates of immobilization. Thus, the application of the isotope pool dilution method with labeled litter proved successful and enabled us to discern contrasting patterns in element immobilization fluxes during the decomposition stages of different leaf litter.

Ecological implications

Cottonwoods are dominant in riparian zones of the western United States, providing more than 80 % of the litter to these streams (Driebe and Whitham 2000). They are often considered foundation species due to their large effects on ecosystem structure and function (Whitham et al. 2006); specifically, the influence of recalcitrant compounds of Populus litter has significant impacts on C and N dynamics within terrestrial ecosystems (Schweitzer et al. 2004, 2008). Our findings extend these ideas, demonstrating that closely related tree species differentially modulate elemental fluxes in streamwater during decomposition, with initial litter characteristics likely driving nutrient cycling during decomposition (Parton et al. 2007). Furthermore, these differences are likely to persist as decomposition progresses, because the relative composition of recalcitrant compounds likely increases due to preferential loss of leachable and more labile compounds during decomposition (Z. Compson, unpublished data).

Plant litter is often not a readily available resource for consumers. Therefore, microbial decomposers establish biogeochemical interactions with their environment, whether streamwater or soil, to supplement deficiencies in C and nutrients, especially when the resource is relatively recalcitrant (Parton et al. 2007). Understanding these biogeochemical interactions with litter should provide insights into nutrient retention, which are responsible for the breakdown, nutrient transfer, and transport of this resource.

Interactions between litter and its environment are especially relevant in forested headwater streams, because these ecosystems are energetically dependent on detrital inputs arriving from the adjacent terrestrial ecosystem (Fisher and Likens 1973; Vannote et al. 1980). Moreover, they are key sites for nutrient retention and transformation along the stream continuum where inorganic N uptake rates often account for more than half of the total input arriving from the watershed (Peterson et al. 2001). Overall, our results indicate that litter characteristics of two cottonwood species drive specific streamwater element requirements of biofilms, suggesting that changes in the proportion of inputs arriving into the streams of these two cottonwood species strongly control stream cycling and export downstream.