Functional and Structural Responses of Hyporheic Biofilms to Varying Sources of Dissolved Organic Matter

Abstract

Headwater streams are tightly connected with the terrestrial milieu from which they receive deliveries of organic matter, often through the hyporheic zone, the transition between groundwater and streamwater. Dissolved organic matter (DOM) from terrestrial sources (that is, allochthonous) enters the hyporheic zone, where it may mix with DOM from in situ production (that is, autochthonous) and where most of the microbial activity takes place. Allochthonous DOM is typically considered resistant to microbial metabolism compared to autochthonous DOM. The composition and functioning of microbial biofilm communities in the hyporheic zone may therefore be controlled by the relative availability of allochthonous and autochthonous DOM, which can have implications for organic matter processing in stream ecosystems. Experimenting with hyporheic biofilms exposed to model allochthonous and autochthonous DOM and using 454 pyrosequencing of the 16S rRNA (targeting the “active” community composition) and of the 16S rRNA gene (targeting the “bulk” community composition), we found that allochthonous DOM may drive shifts in community composition whereas autochthonous DOM seems to affect community composition only transiently. Our results suggest that priority effects based on resource-driven stochasticity shape the community composition in the hyporheic zone. Furthermore, measurements of extracellular enzymatic activities suggest that the additions of allochthonous and autochthonous DOM had no clear effect on the function of the hyporheic biofilms, indicative of functional redundancy. Our findings unravel possible microbial mechanisms that underlie the buffering capacity of the hyporheic zone and that may confer stability to stream ecosystems.

INTRODUCTION

Most headwater streams are net heterotrophic, and allochthonous dissolved organic matter (DOM) is the major subsidy of the heterotrophic metabolism therein (1, 2). It is generally recognized that most of the organic matter processing in streams occurs in the hyporheic zone, which is the part of the streambed where groundwater and streamwater mix (3). The hyporheic zone offers a large surface area for colonization by biofilms, which dominate microbial life in streams and which greatly contribute to DOM retention and transformation (4). High microbial biomass combined with elevated DOM retention renders the hyporheic zone an active compartment of stream ecosystems, even with impacts on large-scale carbon fluxes.

Microbial biofilms in the hyporheic zone encounter allochthonous DOM (AL-DOM) and autochthonous DOM (AU-DOM). For instance, upwelling of shallow groundwater can deliver AL-DOM to streams via the hyporheic zone (3). Furthermore, fresh leaf litter deposited onto the streambed and decaying leaf packs lodged in stream sediments provide a mix of DOM, covering a gradient of bioavailability ranging from labile to semilabile and recalcitrant compounds that can be delivered into the hyporheic zone via downwelling of streamwater. Similarly, AU-DOM from benthic algae can reach the hyporheic zone via downwelling, where it may mix with AL-DOM. In general, such AU-DOM is considered labile to microbial degradation, due to the elevated content of monosaccharides and amino acids (5–7), for instance. On the other hand, AL-DOM, such as leachates from decaying leaf litter, is often aged and depleted in labile compounds and may therefore be resistant (recalcitrant) to microbial degradation (8). The interaction of these different organic matter pools has potentially important implications for carbon cycling in stream ecosystems (9). For instance, labile organic matter may enhance microbial degradation of recalcitrant organic matter through priming or cometabolism (1, 9–11). Therefore, knowledge of the structure and functioning of the microbial communities dwelling in the hyporheic zone and their potential response to variations in the DOM supply is important to better assess the role of the hyporheic zone for biogeochemical processes.

The relationship between environmental controls and community composition and functioning is of central interest for ecology. Both the composition and function of microbial communities have been shown to respond to environmental factors, though it still often remains elusive to what extent changes in community composition are responsible for changes in functioning (12, 13). The manipulation of environmental factors (e.g., DOM quantity and quality) can influence functional and structural community parameters in different ways. Both microbial community structure and function can respond to, or remain unaltered under, environmental influences. Further, the relationship between microbial community composition and functioning may be explained by metabolic plasticity and functional redundancy (14, 15). The former refers to the capacity of a microbial community to adjust its metabolic performance to environmental controls without changing its composition, whereas the latter implies that different taxa can have similar functional roles and the community composition may change while certain functions remain unaltered (14, 15). Furthermore, abundant, dormant, or even dead microorganisms may obscure the relationship between microbial community structure and function. The characterization of microbial community composition by analyses of 16S rRNA bears the potential to circumvent this caveat (16). In fact, RNA has a much shorter life span than DNA and can therefore serve as a tool to capture metabolically active microorganisms.

In this study, we experimentally exposed hyporheic biofilms in bioreactors to AL-DOM extracted from predegraded plant material, simulating allochthonous subsidies to the stream ecosystem. Different amendments simulating AU-DOM sources, including monosaccharides with and without inorganic nutrients (N and P) and algal extracts, were used to mimic the delivery of exudates from benthic algae into the hyporheic zone. A suite of extracellular enzymes was measured as a proxy for microbial function in response to the various DOM sources. We also used 454 pyrosequencing of 16S rRNA and the 16S rRNA gene to study the composition of the bulk and the putatively active microbial communities in the bioreactors. We expected that community composition and extracellular enzyme activities would vary under differing organic substrate conditions. We hypothesized that AU-DOM subsidizing background AL-DOM would affect community composition and function.

MATERIALS AND METHODS

Bioreactor setup and sampling design.

Plug-flow glass bioreactors (17) were used to mimic the sedimentary environment of the hyporheic zone. Sintered glass beads (SIRAN carriers, 2- to 3-mm diameter; Jaeger Biotech Engineering) were used as a substratum for biofilm growth from Oberer Seebach (OSB; Lunz am See, Austria, 600 m above sea level). OSB is a prealpine 3rd-order stream draining a pristine, calcareous catchment (approximately 20 km²) where vegetation is dominated by Fagus sylvatica and Picea abies and, on the gravel bars of OSB, by Salix fragilis. Prior to being packed into the bioreactors, beads were colonized in OSB for 31 days in the dark. This time period was long enough to yield mature microbial communities and typically corresponds to the interstorm interval in OSB. Prior to being packed at equal amounts into bioreactors (n = 25), beads were gently rinsed in filtered streamwater (0.2-μm filter) to remove larger debris and invertebrates. Bioreactors were kept in the dark to avoid growth of phototrophs. Bioreactors were fed with filtered OSB streamwater (0.2-μm filter) and amended with DOM (see below) from 4-liter polypropylene copolymer bottles (Nalgene) and operated in a once-through flow mode using peristaltic pumps (Ecoline VC MS/CA2; Ismatec). Each bioreactor was connected to an individual feed bottle. Bottles were replaced roughly every 48 h and autoclaved between changes. The empty bed contact time in the bioreactors averaged 183 ± 11 min, and the flow rate averaged 0.90 ± 0.05 ml min⁻¹. Experiments were carried out at an average temperature of 18.8 ± 1.3°C.

Experimental treatments and phases.

We designed four experimental phases, lasting a total of 41 days, to mimic pulsed AU-DOM subsidies from algal exudation on top of a continuous AL-DOM input to hyporheic biofilms. In a first phase, all bioreactors received raw streamwater (1.09 mg C liter⁻¹) for 18 days to allow the microbial communities to acclimate to the laboratory conditions. After this acclimation phase, five replicate bioreactors (here termed “starter community”) were destructively sampled, and biofilm was collected to determine extracellular enzymatic activities and microbial biomass (including C and N content) and to extract RNA and DNA for 454 pyrosequencing. In a second phase (lasting 2 days), all remaining 20 bioreactors received sterile filtered streamwater (0.2-μm-filter), containing AL-DOM (0.88 mg C liter⁻¹) produced from willow (see below). In a third phase (the AU-DOM phase, lasting 7 days), we added model AU-DOM in replicates of five bioreactors as follows: glucose (GLC) (0.44 mg C liter⁻¹), glucose and inorganic nutrients (GLC+N+P) (0.44 mg C liter⁻¹, 2,187 μg liter⁻¹ N-NO₃, and 6.82 μg liter⁻¹ P-PO₄), and algal extract (AE) (0.44 mg C liter⁻¹, 2,187 μg liter⁻¹ N-NO₃, and 6.82 μg liter⁻¹ P-PO₄). Five bioreactors receiving streamwater with model AL-DOM, but no AU-DOM, served as control. During this phase, inflow and outflow samples were collected six times for the determination of NO₃, PO₄, O₂, and dissolved organic carbon (DOC) concentrations and three times for the determination of extracellular enzymatic activities. After this AU-DOM addition, we reverted the bioreactors in a fourth phase (the AL-DOM phase, lasting 14 days) to streamwater with AL-DOM (0.88 mg C liter⁻¹). After this final phase, biofilms were sampled from all bioreactors for the analysis of extracellular enzymatic activities and microbial biomass and for 454 pyrosequencing.

Inflow and outflow sampling and solute dynamics.

Streamwater samples for NO₃, PO₄, DOC, and extracellular enzymatic activity analyses were collected from the inflow and outflow of each bioreactor using three-way valves and syringes (100 ml). Outflow samples were collected at the normal flow rate, while inflow samples were collected at a higher rate by diverting the flow for a short time (<5 min). Samples for NO₃ and PO₄ were analyzed on a continuous flow analyzer (FlowSys 3rd generation; SYSTEA Analytical Technologies). Samples for DOC analysis were filtered (Whatman GFF) and DOC concentrations measured on a TOC Analyzer (Sievers 5310C; GE Analytical Instruments). All glassware was acid washed and combusted. The concentration of dissolved oxygen (DO) at the inflow and the outflow was measured using planar optodes in flowthrough cells (PSt3 sensor; Presens, Germany). These measurements were performed 14 times during the experiment.

Oxygen and DOC dynamics.

The accumulated mass (M) of O₂ and DOC removed during the AU-DOM and AL-DOM phases of the experiment was calculated according to the following formula:

$$M={\displaystyle \sum _{p1}^{pn}\left(\Delta C_p\times {\displaystyle {\int }_{t_0}^{t_{\mathrm{int}\hspace{0.17em}}}Qdt}\right)}$$

where ΔC_p is the difference in concentration (DOC or O₂) between inflow and outflow of the bioreactor measured at one sampling point, p₁ and p_n refer to the first and last sampling points of a given phase, Q is the flow rate, and t₀ and t_int represent the duration of an interval encompassing one sampling point. The mass of O₂ and DOC was normalized to the weight of beads contained in each bioreactor.

Production of model AL-DOM.

We produced AL-DOM from a hot-water extraction of crack willow (Salix fragilis), which is a common representative of the riparian vegetation along prealpine streams. Leaf and stem material from S. fragilis was harvested, dried, and ground (Retch MM2) to a fine powder, which was then extracted in MilliQ water at 95°C (60 min). To remove the labile moieties from the willow extract, we subjected the extract to degradation in 15-liter bioreactors with biofilms growing on porous ceramic carriers (Eheim Mech). Bioreactors were continuously aerated to ensure oxic conditions. The decrease in DOC concentration was monitored over 12 days, and we terminated the degradation at day 12 because no significant decrease in DOC concentration was observed during two consecutive days (see Fig. S1 in the supplemental material). At that point, the willow extract was considered relatively resistant to further degradation, and we used this fraction as AL-DOM (filtered on 0.2-μm filters and stored at 4°C). A similar approach was used in previous work (18).

Microbial biomass and cell abundance.

Samples for bulk microbial biomass were harvested from the glass beads by sonication and vortexing, and C and N content was determined using an elemental analyzer (EA1110; CE Instruments, ThermoFisher) (19). An extra sample of glass beads was collected into sterile vials containing 5 ml formaldehyde (2.5%) pending further processing for cell counts. Within 3 weeks after sampling, 20 ml pyrophosphate (0.025 mM pyrophosphate, 2.5% formaldehyde) was added to the glass beads, which were then shaken (60 min) and sonicated three times for 20 s (14% amplitude) to detach the cells from the glass beads. Larger particles were allowed to settle from the supernatant for 20 min. We stained the nucleic acids using SYTOX Green (Life Technologies Corporation) (5 μM final concentration, 15 min), and microbial cells were counted (Quanta; Beckman Coulter) in aliquots.

Extracellular enzymatic activity.

The potential activity of nine extracellular enzymes was measured on the biofilms harvested at the start and at the end of the experiment, as well as in the water from the bioreactor outflow, using substrates linked to aminomethylcoumarin (AMC) and methylumbelliferyl (MUF), respectively, as fluorophores or 3,4-dihydroxyphenylalanine (L-DOPA). All substrates and buffers were purchased from Sigma-Aldrich Chemical Company. The extracellular enzymes were chosen according to their relevance in C, N, and P acquisition (20–23): β-d-glucosidase (EC 3.2.1.21), α-d-glucosidase (EC 3.2.1.20), β-d-xylosidase (EC 3.2.1.37), and cellobiohydrolase (EC 3.2.1.91) are involved in carbohydrate metabolism; leucine-aminopeptidase (EC 3.4.11.1) and endopeptidase (EC 3.4.21-24) are involved in peptide decomposition; phosphatase (EC 3.1.3) is involved in phosphate acquisition; esterase (EC 3.1) breaks ester bonds and serves as a general descriptor of organic matter degradation; and phenol oxidase (EC 1.14.18.1) relates to the hydrolysis of recalcitrant compounds. All fluorogenic substrates were dissolved in 2-ethoxyethanol (Cellosolve), except L-DOPA, which was dissolved in sodium carbonate buffer. The pH of the buffers was set to 7, which approximates the pH of OSB streamwater. MUF and AMC reference standards were prepared with autoclaved MilliQ water. Saturation curves for each enzyme were made to determine the saturation conditions. Glass beads were placed into a sterile Falcon tube (BD Biosciences), and 10 ml of 0.2-μm filtered streamwater from the respective bioreactors was added. Samples were gently sonicated (1 min, 14% amplitude, 1-s pulse; Branson Digital Sonifier 250) to detach and homogenize the microorganisms. We then transferred triplicate aliquots (200 μl) of the biofilm homogenate and the water samples from the bioreactor outflow, respectively, into black 96-well plates (Greiner Bio One) and added 50 μl of the respective substrates. All assays were conducted under substrate saturation conditions and incubated at 19°C in the dark. The fluorescence of the MUF (EX_{365 nm}/EM_{455 nm}) and of the AMC (EX_{364 nm}/EM_{445 nm}) was determined on a TECAN Infinite M200 microplate reader. The absorbance for phenol oxidase activity was measured at 450 nm. Fluorescence and absorbance were repeatedly measured over a period of 0.5 to 12 h on a microplate reader (24). For each enzyme, a negative control for the substrate color (0.2-μm filtered streamwater and the respective substrate) was used to account for abiotic degradation of the substrates. Each 96-well plate also included a reference standard using MUF and AMC in various concentrations. Extracellular enzyme activities were expressed as pmol substrate g⁻¹ beads h⁻¹. To compare extracellular enzyme activities in the water samples among the AU-DOM and the AL-DOM phase, extracellular enzyme activities were integrated over the phase exposure time (25).

Microbial community analysis.

454 pyrosequencing on the 16S rRNA and the 16S rRNA gene served to assess composition and diversity of the putatively active and bulk community, respectively. Total nucleic acids were extracted from the heterotrophic biofilm following the protocol by Urich et al. (26). Briefly, cetyltrimethylammonium bromide (CTAB) buffer and phenol-chloroform-isoamyl alcohol (25:24:1, pH 6.8) were added to a 15-ml lysing matrix E tube (MP Biomedicals LLC) containing 7 g of glass beads with the attached microbial communities. The cells were lysed in a FastPrep machine (MP Biomedicals LLC), followed by nucleic acid precipitation with polyethylene glycol (PEG) 8000. Genomic DNA was digested using the RQ1 RNase-free DNase kit (Promega Corporation) according to the manufacturer's recommendations. An aliquot of the nucleic acid extract was subjected to DNA digestion with RQ1 RNase-free DNase and incubated in a thermal cycler for 30 min at 37°C. Stop solution was added, followed by incubation in a thermal cycler at 65°C for 10 min. The tube was put on ice, and RNA was purified using the MEGAclear kit (Ambion Life Technologies Corporation) according to the manufacturer's recommendations. Absence of DNA was verified by PCR as described below. RNA was transcribed into cDNA by reverse transcription at 42°C for 60 min using RevertAid reverse transcriptase (Thermo Fisher Scientific Inc.) and random hexamer primers (Thermo Fisher Scientific Inc.), followed by an enzyme inactivation step at 70°C for 10 min. Samples without reverse transcriptase and with MilliQ water instead of target RNA served as negative controls. The hypervariable regions V3 and V4 of the bacterial 16S rRNA gene were PCR amplified using the bar-coded forward primer 341F (5′-CCTACGGGNGGCWGCAG-3′) and the reverse primer 805R (5′-GACTACHVGGGTATCTAATCC-3′) (27). For each sample, two different bar codes were used to reduce bar code-specific bias (28). The DNA and cDNA concentrations of the samples were determined using the fluorescent, DNA binding QuantiFluor double-stranded DNA (dsDNA) system kit (Promega Corporation). Equal template concentrations of DNA and cDNA were amplified in all PCRs. Each 25-μl PCR mixture contained each primer at 0.5 mmol liter⁻¹ (Thermo Fisher Scientific Inc.), deoxynucleoside triphosphates (dNTPs) at 0.25 mmol liter⁻¹ (Thermo Fisher Scientific Inc.), bovine serum albumin at 40 mg liter⁻¹ (Thermo Fisher Scientific Inc.), MgCl₂ at 2.5 mmol liter⁻¹ (Thermo Fisher Scientific Inc.), and 1 U Taq-DNA polymerase with the recommended PCR buffer (Thermo Fisher Scientific Inc.). Samples were amplified using an initial denaturing step at 94°C for 2 min, followed by 25 cycles of denaturation at 94°C for 30 s, annealing for 30 s starting with a 62°C annealing temperature and decreasing to 50°C (reduction of 0.5°C per cycle), elongation at 72°C for 1 min, and a final elongation at 72°C for 10 min. Each PCR included a negative control. PCR products were run on a 1.5% agarose gel (Top Vision Agarose; Thermo Fisher Scientific Inc.). The PCR bands were cut from the gel and purified using the gel extraction kit (Qiagen) according to the manufacturer's recommendations. The purified PCR products were quantified on a 1.5% agarose gel using the Gel Doc XR+System (Bio-Rad Laboratories, Hercules, CA, USA) in combination with the MassRuler DNA Ladder Mix and the recommended loading dye (Thermo Fisher Scientific Inc.). The amplicon concentrations obtained by gel quantification were verified by using the DNA binding QuantiFluor dsDNA system kit (Promega Corporation). Amplicons were pooled in equimolar concentrations and sequenced on a GS FLX titanium sequencer at the Center for Genomic Research (University of Liverpool, Liverpool, United Kingdom).

Statistical analyses.

All statistical analyses were performed using the software and statistical computing environment R (R Development Core Team, 2013). Significant differences in removal of oxygen and DOC between the treatments and the control were tested using ANOVA and post hoc Tukey's test. Extracellular enzymatic activities were tested using the nonparametric Kruskal-Wallis test and the Mann-Whitney U test for pairwise comparisons (P < 0.05). The significance value was adjusted for multiple comparisons using the Bonferroni correction. A Mantel test and partial Mantel's test with Spearman's rank correlation with 999 permutations were performed to test correlations between environmental variables (NO₃, PO₄, DOC, and O₂ concentrations at the inflow of the bioreactors), species composition, and the respective extracellular enzyme rates.

The 454 pyrosequencing data were denoised, and reads were clustered at a 97% identity level to operational taxonomic units (OTUs) using the software package AmpliconNoise V1.28 (29). Taxonomic assignments were determined using the CREST classifier (30). Alpha diversity was calculated as richness and the number equivalents of the Shannon and of the Gini-Simpson indices, which differently weigh abundant and rare species (31). All samples were rarefied to the lowest number of reads obtained from a sample (2,331 reads) prior to analysis. We computed similarity matrices using the Horn index (32) from randomly resampled 454 pyrosequencing data (33) and from the extracellular enzyme activities, respectively. These similarity matrices were subjected to multidimensional scaling (MDS) analysis to visualize community and functional dynamics of the biofilm bacteria. Permutational multivariate ANOVA (PERMANOVA) was used to test significant differences among the treatments. Alpha diversity measures were tested for significant differences between treatments using ANOVA and post hoc Tukey's test. Differences in beta diversity (variability in community composition between bioreactors) were assessed by calculating the distance to centroid of each treatment in multidimensional space (34, 35).

Nucleotide sequence accession numbers.

The following accession numbers were obtained upon submission of the sequences to the GenBank database: SRX433107 and SRX462313.

RESULTS

Dynamics of dissolved organic carbon and oxygen in the bioreactors.

DOC removal was significantly lower in the control, the GLC, and in the algae (AE) treatments during the AL-DOM phase (P < 0.05) than during the AU-DOM phase (Fig. 1). DOC removal was significantly lower in the control treatment than in the GLC and the AE treatments during the AU-DOM phase (P < 0.05), whereas DOC removal did not differ in the AL-DOM phase among treatments (P = 0.07). The removal of oxygen in the bioreactors was significantly lower in the control treatment than in the AE treatment during the AU-DOM phase (P < 0.01), while it did not differ among treatments during the AL-DOM phase (P = 0.24).

FIG 1.

Temporal dynamics of DOC and O₂ removal cumulated over the AU-DOM phase (white) and the AL-DOM phase (black) of the different treatments. Treatments are marked as follows in all graphs: S, starter community; G, GLC; GNP, GLC+N+P; AE, algae; C, control. Outliers are displayed as dots. A hash tag (#) indicates significant differences (P < 0.05) between the control and the AU-DOM treatments (G, GNP, and AE).

Extracellular enzyme activities in the bioreactor outflow.

Phenol oxidase activity was significantly lower (P < 0.001) in the control, the GLC, the GLC+N+P, and the AE treatments during the AL-DOM phase than during the AU-DOM phase (Fig. 2). Esterase, phosphatase, and leucine-aminopeptidase activities did not show significant differences between the control, the GLC, the GLC+N+P, and the AE treatments during the AU-DOM and the AL-DOM phase, respectively (P = 0.99). Endopeptidase activity was significantly lower (P < 0.05) in the control treatment during the AL-DOM phase than during the AU-DOM phase. Endopeptidase activity was significantly higher in the control than in the AE treatment during the AU-DOM phase (P < 0.05), whereas endopeptidase activity did not significantly (P = 0.13) differ among treatments during the AL-DOM phase.

FIG 2.

Extracellular enzyme activity (EEAs) of 5 different enzymes cumulated over the AU-DOM phase (white) and the AL-DOM phase (black) of the different treatments (see Fig. 1 legend for abbreviations and significance labeling). Displayed are phenol oxidase activity (A), leucine-aminopeptidase activity (B), esterase activity (C), endopeptidase activity (D), and phosphatase activity (E) as nmol substrate converted h⁻¹.

Microbial biomass.

Abundance of microbial cells attached to the beads was significantly lower in the starter community than in the GLC, the GLC+N+P, the AE, and the control treatments (P < 0.001) (Table 1). Biomass (total C and N content) of the biofilms did not show significant differences among treatments (P = 0.81).

TABLE 1.

Cell abundance and total C and N content of the biofilm biomass^a

Sample	No. of cells (104)	C content (mg)	N content (mg)
Starter community	38.75 ± 5.60	0.072 ± 0.02	0.008 ± 0.002
GLC	79.40 ± 11.30*	0.114 ± 0.03	0.013 ± 0.005
GLC+N+P	71.40 ± 10.21*	0.103 ± 0.02	0.012 ± 0.003
Algae	66.20 ± 11.80*	0.092 ± 0.01	0.011 ± 0.001
Control	70.44 ± 9.00*	0.083 ± 0.02	0.009 ± 0.002

^aValues are means ± standard deviations per g of beads. *, significant difference (P < 0.05) between the starter community and the experimental treatments.

Community composition and diversity.

The nonmetric MDS (NMDS) revealed a clear separation between bulk and active community compositions (P < 0.001) (Fig. 3). The starter community differed significantly in both bulk and active community compositions from the GLC, the GLC+N+P, the AE, and the control treatments (P < 0.001). The community composition of the control treatment did not differ from the GLC, the GLC+N+P, and the AE treatments in the bulk (P = 0.12) and in the active (P = 0.77) communities. Beta diversity (i.e., distance to centroid in NMDS space) of the active starter community was significantly higher than that of the GLC, GLC+N+P, and AE treatments (P < 0.05) (Fig. 4). Furthermore, beta diversity of the control treatment was significantly lower in the active community than that of the GLC and the GLC+N+P treatments (P < 0.05). The beta diversity of the bulk starter community was significantly higher than that from the AE treatment (P < 0.05) (Fig. 4).

FIG 3.

Nonmetric multidimensional scaling (NMDS) ordination based on the Horn distance of the bulk (triangles) and the active community composition (circles) of each bioreactor from the different treatments (see Fig. 1 legend for abbreviations). Symbols are grouped with an ordiellipse (95% confidence interval) to illustrate how treatments cluster in NMDS space.

FIG 4.

Beta diversity for the active (left panel, white) and the bulk (right panel, black) community compositions under the different treatments. See Fig. 1 legend for abbreviations; *, significant difference (P < 0.05) between the starter community and the treatments (G, GNP, AE, and C); #, significant differences (P < 0.05) between the control and the AU-DOM treatments (G, GNP, and AE). Outliers are displayed as dots.

Diversity was generally high in the bioreactors. The starter community showed significantly higher richness (as OTUs) than the GLC, the AE, and the control treatments in the active community (P < 0.05), whereas the starter community showed significantly higher richness than the GLC, the GLC+N+P, and the AE treatments (P < 0.05) in the bulk community. Evenness of the active communities was significantly higher in the starter community than in the AE treatments and the control (P < 0.05), whereas no difference in evenness of bulk communities was detected among treatments. Similar trends were observed when comparing the Shannon and Gini-Simpson diversity indices for active and bulk communities between the starter community and the treatments (Table 2).

TABLE 2.

Characteristics of bulk and active communities under different treatments^a

Sample	Gini-Simpson NE	Shannon index NE	Richness	Evenness
Bulk community
Starter community	265.5 ± 22.2	555.0 ± 23.7	994.1 ± 24.3	0.92 ± 0.004
GLC	232.3 ± 28.4	510.3 ± 23.3	920.5 ± 19.6*	0.91 ± 0.005
GLC+N+P	259.4 ± 30.5	528.8 ± 31.7	928.9 ± 30.2*	0.92 ± 0.005
Algae	215.0 ± 36.4	494.3 ± 23.4*	919.0 ± 26.5*	0.91 ± 0.006
Control	234.1 ± 54.4	525.8 ± 44.6	953.0 ± 35.3	0.91 ± 0.008
Active community
Starter community	207.0 ± 9.6	481.6 ± 11.0	946.3 ± 12.1	0.90 ± 0.002
GLC	79.9 ± 23.9*	356.0 ± 68.8*	869.5 ± 55.0*	0.87 ± 0.026
GLC+N+P	97.2 ± 51.1*	383.2 ± 100.5	875.7 ± 57.2	0.87 ± 0.034
Algae	63.8 ± 15.9*	315.0 ± 38.3*	821.6 ± 35.7*	0.86 ± 0.015*
Control	75.3 ± 18.9*	327.4 ± 44.4*	830.9 ± 18.1*	0.86 ± 0.018*

^aValues are means ± standard deviations. NE, number equivalents; *, significant difference (P < 0.05) between the starter community and the experimental treatments.

Overall, we detected sequences from 50 different phyla. The most abundant phylum in the bulk and the active communities was Proteobacteria (40.1% ± 2.4% and 44.4% ± 4.8%, respectively). Several of the bacterial phyla displayed shifts in abundance from the starter community to the experimental treatments that were sampled at the end of the experiment. In both the bulk and the active communities, the relative abundance of Betaproteobacteria and Planctomycetes was significantly (P < 0.05) lower in the starter community than in the treatments, whereas Verrucomicrobia and Acidobacteria increased in abundance. In addition, a significant (P < 0.05) decrease in abundance for Bacteriodetes and an increase for Chloroflexi could be detected in the bulk community but not in the active community. There was no significant difference in abundance of any phylum between the AU-DOM treatments and the control treatment. At the genus level, several taxa displayed shifts in the active and the bulk communities between the starter community and the various treatments (see Fig. S2 in the supplemental material). For instance, Prosthecobacter (Verrucomicrobia) was the most abundantly assigned genus in our data set and was relatively more abundant in the active than in the bulk community; Gemmata and Hirschia did not follow that pattern (Fig. 5). Our data show that Gemmata exhibited higher relative abundance upon AU-DOM and notably upon algal extract amendments; Planctomyces showed an inverse pattern (Fig. 5).

FIG 5.

Taxonomic affiliation of the active (A) and the bulk (B) community composition under the different treatments as determined by 454 pyrosequencing. See Fig. 1 legend for abbreviations. Asterisks above the bars indicate a significant difference (P < 0.05) between the starter community and the experimental treatments. The most abundant phyla and 3 specific genera are displayed. Unassigned sequences are not shown here and accounted for less than 1% of relative abundance at the phylum level and for approximately 60% at the genus level.

Extracellular enzyme activities of the hyporheic biofilm community.

Phosphatase and endopeptidase displayed significantly higher activities in the starter community than in all treatments and in the control (P < 0.05) (Fig. 6). However, phosphatase and endopeptidase activities did not differ between the control and the treatments (P = 0.88). The activities of the other extracellular enzymes did not show significant differences among treatments (P = 0.70).

FIG 6.

Extracellular enzyme activities of 9 different enzymes from the hyporheic biofilm communities (see Fig. 1 legend for abbreviations; *, P < 0.05). Displayed are phenol oxidase activity (A), xylosidase activity (B), cellobiosidase activity (C), α-glucosidase activity (D), β-glucosidase activity (E), endopeptidase activity (F), leucine-aminopeptidase activity (G), phosphatase activity (H), and esterase activity (I) as pmol substrate converted g⁻¹ beads h⁻¹. Outliers are displayed as dots.

All treatments displayed similar enzyme activities in NMDS space (P = 0.56) (Fig. 7). The starter community overlapped the GLC, the GLC+N+P, the AE, and the control treatments. This indicates a higher overlap on the functional level (Fig. 7) than observed for the community composition (Fig. 3). A Mantel's test showed that environmental variables (NO₃, PO₄, DOC, and O₂ concentrations in the inflow) were significantly correlated with the composition of the active (R = 0.64, P < 0.001) and the bulk (R = 0.64, P < 0.001) communities. The active community composition significantly correlated with the extracellular enzyme activities (R = 0.23, P < 0.05). This pattern was consistent when a partial Mantel's test was performed to test for the correlation between the active community composition and the enzymatic activity, while controlling for the effect of the environmental variables (R = 0.23, P < 0.05). On the other hand, neither the environmental variables nor the bulk community composition showed any correlation with the extracellular enzyme rates (R = 0.15, P = 0.18). The active (R = 0.17, P = 0.13) and the bulk (R = 0.06, P = 0.35) community compositions did not significantly correlate with the extracellular enzyme activities and the environmental variables when the starter community was omitted from the analysis.

FIG 7.

NMDS ordination based on Horn distance of 9 extracellular enzyme activities from each bioreactor of the different treatments (see Fig. 1 legend for abbreviations). Symbols are grouped with an ordiellipse (95% confidence interval) to illustrate how treatments cluster in NMDS space.

DISCUSSION

Hyporheic microorganisms receive AL-DOM, which has often undergone degradation and transformation before entering streams and is therefore considered less labile than AU-DOM (36, 37). The degradation of this AL-DOM may greatly contribute to carbon cycling of streams and may even have potential impact on large-scale carbon fluxes (3). However, little is known on the microbial communities that are involved in these biogeochemical processes and their underlying mechanisms. Our study illuminates the relationship between biofilm structure-function coupling and the processing of AU-DOM and AL-DOM in bioreactors mimicking the hyporheic zone.

Despite the increased metabolism upon AU-DOM additions, there were no corresponding changes in extracellular enzyme activities measured in the outflow of the bioreactors. We expected that AU-DOM provides easily available energy that would enable microorganisms to express more extracellular enzymes enhancing AL-DOM degradation (9). The tendency of higher esterase and endopeptidase activities in the AU-DOM treatments indicates a delayed expression increase of these enzymes. However, as these differences were not significant and no similar effect was observed for these enzymes on the biofilm biomass, priming may not be of relevance in the hyporheic zone. This is consistent with a parallel study on carbon fluxes in the same system (41).

We expected that varying DOM sources differentially affect community composition of hyporheic biofilms. No obvious effect of the autochthonous pulse could be detected on community composition of either the active or the bulk community in response to AU-DOM additions. There may be several explanations for this. Although all DOM additions (including glucose) represented a manipulation within the range of seasonal variation in the OSB streamwater, it might not have been pronounced enough to induce a diversion between treatments and the control. It is likely that concentrations within or directly adjacent to biofilms can be considerably higher than in the streamwater, for example, during a benthic algal bloom. Also, we recognize that we can compare community composition only after the AL-DOM phase and that we cannot draw any conclusions on community dynamics prior to that. It is possible that a shift in community composition occurred directly after the AU-DOM pulse and that the communities converged again to a more similar state until we sampled them at the end of the AL-DOM phase. Another possibility for the relatively minor effect of the AU-DOM pulse on community composition could be that labile DOM is easily metabolized without prior enzymatic degradation by most microbes, thereby not necessitating a change in community composition for an efficient exploitation of these resources.

The importance of AL-DOM for aquatic microorganisms has been shown previously; it is generally believed that AL-DOM provides a relatively continuous energy source that supports slow but steady microbial growth independent of AU-DOM pulses during algal blooms, for instance (37). The consistent divergence in community composition from the starter community that we observed in all treatments is likely attributable to the additions of AL-DOM. We suggest that AL-DOM rather than AU-DOM drives the community dynamics in hyporheic biofilms.

Beta diversity can be related to ecosystem productivity when stochastic community assembly driven by priority effects leads to a higher variability of community composition in productive systems (38). Underlying this notion is the fact that efficient colonizers outcompete others and that various species can thrive in productive ecosystems where community assembly becomes therefore random. We propose that an elevated beta diversity of the active community after the AU-DOM addition is possibly related to such priority effects. As microorganisms encounter presumably labile AU-DOM, stochasticity may drive community assembly, and those taxa that are able to quickly exploit the new resource soon become abundant. Due to priority effects, different microorganisms grow abundant in the different bioreactors, leading to the observed variation in composition. This assumption is supported by elevated beta diversity paralleling increased DOC and oxygen removal as proxies for biofilm metabolism.

We found that the composition between the bulk and the active communities differed markedly, which is likely attributable to various contributions from microorganisms with differing physiological states (39, 42). This may also reflect the varying provenience of microorganisms mixing from various terrestrial and aquatic (including groundwater) habitats in the hyporheic zone. Both the active and bulk communities depicted clear compositional shifts at the phylum level between the starter communities and the treatments. This is notable, as it demonstrates an ecological response even at a higher taxonomic rank (43) to relatively small environmental changes. Obviously, the Proteobacteria were the most responsive phylum, with the genus Hirschia from the Alphaproteobacteria providing a good example for a remarkable shift in relative abundance. Hirschia was reported to degrade various monosaccharides, amino acids, sugar alcohols, and even cellobiose as suggested by elevated β-glucosidase activity (40). These compounds generally characterize autochthonous and allochthonous DOM, respectively, in aquatic ecosystems and assumedly also in the hyporheic zone, where they mix.

We found different responses of community composition and function (that is, extracellular enzymatic activities) of the hyporheic biofilms to AU-DOM amendments. This is indicative of functional redundancy as has been previously reported from freshwater ecosystems, yet with differing levels of relevance (44, 48); Frossard et al. (45) even reported a clear disconnect between microbial community structure and enzymatic activities. We further investigated the structure-function coupling of hyporheic biofilms by exploring the relationship between bulk or active community and extracellular enzymatic activities. We expected that the environment could influence the microbial community composition with consequences for the extracellular enzymatic activities. The fact that we found a relationship between the active community composition and the respective extracellular enzyme activities, while controlling for the effect of the environment, underlines the link between the active community and enzyme expression. This seems reasonable, given that transient physiological states of microorganisms are considered to be a mechanism that contributes to the buffering against environmental fluctuations, including resource availability, and to the maintenance of microbial diversity and functioning (46, 47). Few studies have addressed structure-function coupling of hyporheic microorganisms (12, 19), which our study expands now by showing the apparent need to consider the active rather than the bulk community. No clear relationship between community structure and function existed when the starter community was omitted from the analysis, indicating that the observed relationship is driven by the temporal dimension in our experiment.

It is commonly believed that the hyporheic zone is buffered against environmental fluctuations and that it may contribute to the stability of the ecological and biogeochemical processes of the stream ecosystem (3). Our experimental work unravels microbial mechanisms that potentially contribute to the buffering capacity of the hyporheic zone. For instance, a high taxon turnover (that is, beta diversity) and related priority effects and stochastic community assembly are beneficial in an ecosystem that is characterized by unpredictable hydrology and subsidies of DOM differing in composition and bioavailability. Furthermore, functional redundancy may ensure the continuous provision of essential ecosystem functions independent of community composition, which itself depends on varying environmental processes. Finally, the pool of putatively nonactive microorganisms can sustain the active community as environmental conditions change to favor such microorganisms, which in turn may then upon reactivation sustain critical functions in stream ecosystems.