Introduction

Though wetlands account for only ~10 % of terrestrial land area (Zedler and Kercher 2005), their influence on the global carbon (C) cycle is disproportionately large. For example, wetland soils store 45–70 % of terrestrial organic C (Mitra et al. 2005) and are responsible for nearly 25 % of global methane (CH4) emissions (Conrad 2009). Carbon sequestration is enhanced in these environments due to a combination of high primary productivity and slow rates of decomposition. The quality and availability of organic material (OM), as well as an interacting suite of environmental factors (e.g., soil moisture and pH), determine the degree of OM storage versus mineralization to carbon dioxide (CO2) and/or CH4 (Segers 1998; Megonigal et al. 2004; Kayranli et al. 2010). Understanding how OM properties affect the balance between sequestration and mineralization is particularly relevant in the context of wetland restoration and creation, as these activities often involve OM amendments to hasten the development of organic-rich reduced soils (Davis 1995; Mitsch and Gosselink 2000).

The OM transformations that result in the production of CH4 and/or CO2 are driven by diverse microbial communities that depend on extracellular enzyme activity (EEA) to breakdown complex organic polymers into soluble compounds that can be transported into the cell and metabolized. Consequently, this depolymerization is the putative rate-limiting step in OM decomposition (Sinsabaugh et al. 1991) and enzyme depolymerization rates have been associated with microbial respiration in aquatic ecosystems (Arnosti and Holmer 2003; Arnosti and Jørgensen 2006; Baltar et al. 2009). In soils however, surprisingly few studies have explored the relationship between EEA and either C mineralization rates (Freeman et al. 1997, 2001; Allison and Vitousek 2005) or the composition of the associated microbial communities (Kourtev et al. 2003; Gallo et al. 2004; Costa et al. 2007; Kaiser et al. 2010). Even less consideration has been given to how these three components interact. Though some studies have found a strong relationship between microbial community composition and activity (Cleveland et al. 2007; Li et al. 2011; Goberna et al. 2012; Lazar et al. 2012), many others have not (Galand et al. 2003; Liu et al. 2011; Fromin et al. 2012). One common hypothesis for this inconsistency is that the high functional redundancy within microbial communities limits our detection of community structure–function relationships, especially when considering processes with a relatively ubiquitous distribution among taxa (e.g., respiration to CO2, Griffiths et al. 2000; Nannipieri et al. 2003). Some have proposed that community structure only regulates “narrowly” distributed functions (Schimel 1995; McGuire and Treseder 2010), i.e., ones that are performed by only a small group of organisms with specialized physiological pathways such as methanogens.

In this study, we examined how microbial community structure and EEA regulate greenhouse gas (CO2 and CH4) production in wetlands receiving long-term in situ soil amendments of either plant litter or compost. This research was conducted in a recently restored tidal freshwater wetland, and the results have implications for recovering ecosystem services facilitated by OM-rich soils in impaired wetlands while simultaneously minimizing the production of CH4.

Methods

Experimental design

In January 2011, soil (5–15 cm depth) was collected from the middle of a 30-ha tidal freshwater wetland at Virginia Commonwealth University’s Walter and Inger Rice Center for Environmental Life Sciences (James River, Charles City County, VA, USA; 37°20′05″N, 77°12′27″W). This wetland was an impounded lake for nearly 70 years before a storm breached the dam and restored natural wetland hydrology to the site in 2006. The site was continually saturated, usually with standing water on the surface, and dominated by obligate wetland vegetation such as Leersia oryzoides, Juncus effusus, and Nuphar luteum. The soil had an OM content of 8 %, a C:N ratio of 10 (by mass), a pH of 6.4, and soil texture was 30 % sand, 55 % silt, and 15 % clay. Following extensive homogenization in the laboratory, one fraction of the collected soil was amended with plant litter (standing dead material of the above mentioned species harvested from the field site in early December 2010; 99 % OM content, C:N = 72). A second fraction was amended with compost (Black Kow, Oxford, FL; organic blend containing 26 % OM, C:N = 18). The compost and litter amendments were similar in particle size (0.1–5 mm in diameter). Amendments were added to raise the soil OM content to approximately double ambient levels. A third soil fraction was unamended and served as an experimental control.

Litter bags (12 cm × 22 cm), constructed of 0.5-mm Nitex mesh (Wildlife Supply Company, Buffalo, NY, USA), were filled with 400 ml (~215 g dry weight) of control or amended soil for in situ incubation. In January 2011, 10 bags of each type were buried (5–15 cm depth) at random locations within a single 20 × 20 m experimental plot near the soil collection site. Samples were incubated for either 6 or 18 months (until July 2011 and 2012, respectively). At each sampling event, five replicate bags of each type were collected, as were five intact field cores. Samples were placed in airtight plastic bags, quickly transported back to the laboratory, and subdivided for soil characterization (200 g) and molecular genetic analyses (5 g, immediately archived at −20 °C). For the 18-month sampling, subsamples were also removed for determination of CO2 and CH4 production (40 g, stored for 7 days at 4 °C) and analysis of extracellular enzyme activities (10 g, stored up to 5 days at 4 °C). The field cores were analyzed only for soil properties and gas production rates, and served to provide context for interpreting the experimental manipulations.

Environmental analyses

At both the 6- and 18-month sampling events, soil redox potential and pH were measured using a Hanna Combo pH and ORP probe (QA Supplies, Norfolk, VA, USA). Soil moisture (%) was determined gravimetrically (100 ± 5 °C for 72 h), and OM (%) was measured as the mass loss on ignition following combustion at 500 °C for 4 h. Total carbon and nitrogen contents were determined using a Perkin Elmer CHNS/O Analyzer (Waltham, MA, USA) following grinding and acidification of samples using 10 % hydrochloric acid.

In addition, at the 18-month sampling, porewater was extracted and analyzed for dissolved nutrient concentrations. Briefly, water was collected from 50-ml soil samples by centrifugation (3,000×g for 15 min), filtered using a 0.45 μm pore-size mixed cellulose ester syringe filter, and stored at −20 °C until it could be analyzed for: (i) ammonium (NH4 +) using the indophenol colorimetric assay of Grasshoff et al. (1983), (ii) dissolved organic carbon (DOC) using a Shimazdu TOC analyzer (Columbia, MD, USA), (iii) total dissolved nitrogen (TDN) and phosphorus (TDP) using a Skalar Sans Plus System (Buford, GA, USA).

Molecular analyses

Whole-community DNA was extracted from 0.5-g subsamples of soil using the MoBio PowerSoil DNA Isolation Kit (Carlsbad, CA, USA) and stored at −20 °C. DNA purity and concentration were analyzed using a Nanodrop ND-1000 (Thermo Scientific, Wilmington, DE, USA). All DNA extracts and PCR products were verified using agarose gel (1.5 %) electrophoresis and ethidium bromide staining.

Microbial gene abundance via qPCR

Quantitative polymerase chain reaction (qPCR) assays were performed to assess the genetic potential of the microbial communities (Smith and Osborn 2009). Assays were performed using SsoAdvanced SYBR Green qPCR Supermix (BioRad, Hercules, CA, USA) and a BioRad CFX 96 Real-Time System; data were analyzed using Bio-Rad CFX Manager Version 2.1. Results were reported as the log(10) of the number of gene copies per gram of OM after averaging three technical replicates per sample and comparing to appropriate standard curves.

To estimate total bacterial abundance, the primers Eub338 and Eub517 were used to target the 16S rRNA gene (Fierer et al. 2005). Genomic DNA from Escherichia coli (Strain 11775, ATCC, Manassas, VA, USA) was used to establish the standard curve (average efficiency = 101 %, r2 = 0.99). Reactions (20 μl) were performed with 1.2 ng DNA template and 0.1 μM concentrations of each primer; thermal cycling conditions were: 95 °C for 4 min, and 40 cycles of 30 s at 95 °C, 30 s at 55.5 °C, and 60 s at 72 °C. The abundance of archaea was estimated using the primers Arch 967F and Arch-1060R (Karlsson et al. 2012), again targeting the 16S rRNA gene. Standard curves (average efficiency = 94 %, r2 = 0.99) used genomic DNA from Methanococcus voltae (Strain A3, ATCC). Reactions (20 μl) used 2 ng DNA template and 0.3 μM primers; thermal cycling conditions were: 95 °C for 5 min, and 40 cycles of 20 s at 95 °C, 20 s at 59 °C, and 30 s at 72 °C. Finally, methanogen abundance was estimated using the mlas and mcrA-rev primers to target the methyl coenzyme-M reductase encoding mcrA functional gene (Steinberg and Regan 2009). As with archaea, Methanococcus voltae genomic DNA was used for the standard curve (average efficiency = 92 %, r2 = 0.99). Reactions (20 μl) had 2 ng DNA template and 0.56 μM mlas and 0.70 μM mcrA-rev primer concentrations; thermal cycling conditions were: 95 °C for 5 min, and 50 cycles of 20 s at 95 °C, 20 s at 59 °C, and 45 s at 72 °C.

Community structure via T-RFLP

Microbial community structure was analyzed using terminal restriction fragment length polymorphism (T-RFLP) targeting the 16S rRNA gene of bacteria and archaea and the functional gene mcrA for methanogens. All PCR reactions (50 μl) were performed with 10 mM Tris HCl (pH 8.3), 50 mM KCl, 200 μM of each dNTP, 20 μg BSA, and 2.5 units of AmpliTaq DNA polymerase (reagents obtained from Applied Biosystems, Foster City, CA, USA). Bacteria PCRs including the domain-specific primers 27F (labeled with FAM) and 1492R at a concentration of 0.2 μM each (Lane 1991), 3.0 μM MgCl2, and 1.2 ng DNA. Thermal cycling conditions were: 95 °C for 3 min, 30 cycles of 45 s at 95 °C, 60 s at 57 °C, 120 s at 72 °C, followed by 72 °C for 7 min (PTC-100 Thermal Controller, MJ Research, Waltham, MA, USA). Archaea PCRs included the primers 21F (labeled with FAM) and 958R (Cytryn et al. 2000), each at 0.2 μM, as well as 1.5 μM MgCl2, and 4 ng DNA. Thermal cycling conditions were: 94 °C for 3 min, 34 cycles of 60 s at 95 °C, 60 s at 55 °C, 60 s at 72 °C, followed by 72 °C for 7 min. The methanogen mcrA gene was targeted with MLf (labeled with FAM) and MLr (Smith et al. 2007) in a reaction with 60 nM of each primer, 2 μM MgCl2, and 4–8 ng DNA. Touchdown PCR thermal cycling conditions were: 95 °C for 3 min, six cycles of 45 s at 95 °C, 60 s at 56 °C (−0.5 °C cycle−1), 60 s at 72 °C, followed by 31 cycles of 45 s at 95 °C, 60 s at 53 °C, 60 s at 72 °C and final extension at 72 °C for 7 min.

PCR products were purified using the MinElute 96 UF PCR purification kit (Qiagen, Valencia, CA, USA) prior to restriction enzyme digest (associated materials from New England Biolabs, Ipswich, MA, USA). Digests were conducted in 1× Buffer #4 with 30 ng BSA, using either 10 units of Hha1 (16S rRNA) or 20 units of RsaI (mcrA). After digestion (16 h at 37 °C, 20 min at 65 °C), amplicons were purified using the MinElute kit, recovered in molecular-grade water, and detected using capillary electrophoresis with a MegaBACE 1000 DNA Analysis System (Amersham Biosciences, Buckinghamshire, UK). An aliquot of 50–120 ng of purified, digested PCR product was combined with 0.3 μl of MapMarker 400 ROX ladder (Bioventures, Murfreesboro, TN, USA) plus 4.75 μl injection buffer (0.1 % Tween-20). Samples were injected at 3 kV for 100 s, and electrophoresed using genotyping filter set 1 for 100 min at 10 kV. T-RFLP fragments between 70 and 400 base pairs (bp) were analyzed using Fragment Profiler software (Version 1.2; Amersham Biosciences) using a 1 bp size differential and a 15 relative fluorescent unit peak height threshold. Samples were standardized by calculating peak area as a percent of the total sample fluorescence; peaks accounting for <1 % of total sample fluorescence were removed prior to analysis.

Extracellular enzyme activity (EEA)

Soil slurries were prepared fresh each day of analysis by sonicating 1.0 g soil in 100 ml sterile deionized water (15 W for 2 min; Misonix Sonicator 3000, Farmingdale, NY, USA). The slurries were kept on a shaker table (150 rpm) during use to prevent settling of the soil particles. The activities of five hydrolytic enzymes were measured using fluorometric assays following modified protocols from Stursova et al. (2006) and German et al. (2011) with reagents obtained from Sigma-Aldrich Co. Ltd. (Table 1). Three technical replicates of each sample were assayed at each of ten substrate concentrations, as were three negative (no sample) controls. For the methylumbelliferone (MUB) assays, MES buffer (0.1 M, pH 6.1) was used, and quench curves were established for each sample using a range from 0 to 9 nmol MUB. For the amino-4-methylcoumarin hydrochloride (AMC) assay, a Trisma buffer was used (50 mM, pH 7.8) and quench curves ranged from 0 to 7.5 nmol AMC.

Table 1 Summary of enzymes assays used in this study, their natural substrates and products, as well as artificial substrates and concentration ranges used for determination of enzyme kinetics
Full size table

Plates were prepared by adding soil slurry (50 μl) first, followed by substrate, and then buffer sufficient to achieve a final volume of 200 μl per well. Plates were pre-incubated at 30 °C for either 1 h (for CBH, LAP, and AP) or 4 h (for BG and BX), and then read for an additional 6 h using a Synergy 2 plate reader (Biotek, Winooski, VT, USA) programmed for 360 nm excitation and 460 nm emission wavelengths and an incubation temperature of 30 °C. Activity was calculated for each sample after fitting a regression of the fluorescent reading versus MUB or AMC concentration for the corresponding quench curve. Rates were determined as the change in MUB or AMC generated in each sample during the 6 h incubation in the plate reader (each plate was read a minimum of twelve times). Technical replicates were averaged, and maximum reaction velocity (Vmax) and half-saturation constant (Km) values were calculated on Sigma Plot Version 10 (San Jose, CA, USA) using the Michaelis–Menten hyperbola function in the regression wizard.

Phenol oxidase (POX) activity was measured colorimetrically (Sinsabaugh et al. 2003). Three technical replicates containing soil slurry (50 μl), 50 mM sodium bicarbonate buffer (pH 6.1), and l-DOPA (6.5 mM) were performed for each sample, as were triplicate no-sample and no l-DOPA controls. The plates were incubated in the dark at 30 °C for 30 min and then read on the Synergy 2 at 460 nm wavelength for 6 h at 30 °C.

Anaerobic CO2 and CH4 Production

Production of CO2 and CH4 was measured using an anaerobic slurry assay (Neubauer et al. 2005). Briefly, homogenized soil samples (7.0 ± 0.2 g) were combined with 7 ml of filtered (glass microfiber filter GF/C; Whatman Piscataway, NJ, USA), deoxygenated porewater in a 125-ml serum bottle under anaerobic conditions using an N2-filled glove bag. Two technical replicates were prepared for each sample, and slurries were pre-incubated overnight (~16 h at room temperature, 23 °C). The next morning, the headspace in each bottle was thoroughly flushed with N2 to initiate a two-day experimental incubation. Gas samples (5 ml) were obtained from the headspace at 0, 8, 22, 32, and 46 h by shaking the slurry briefly, injecting 5 ml of N2, and immediately withdrawing 5 ml of gas. Measuring headspace gas concentrations will underestimate potential production rates to the extent that gases accumulate in the slurry liquid rather than in the headspace, although this will not affect the relative comparison between our treatments since experimental conditions (pH, salinity, volumes of liquid and headspace) were similar in all bottles. Concentrations of CO2 were measured on a LI-COR LI-7000 infrared gas analyzer (Lincoln, NE, USA), and CH4 was measured on a Shimadzu GC-14A gas chromatograph with flame ionization detector. All samples showed a linear increase in gas concentration over time, and production rates (nmol CO2 or CH4 produced per g of OM per hour) were calculated using linear regression. Median correlation coefficients were 0.97 for CO2 and 0.99 for CH4. Analytical precision was ±0.87 % for CO2 and CH4 (mean coefficient of variation for replicate injections of CO2 and CH4 standards).

Statistical analyses

By design, our treatments differed in their OM content. To account for this in data analysis, we normalized all microbial abundance, enzyme activity, and gas production data per gram of OM; this allowed us to focus on the effect of OM type without the confounding effect of amount. Shapiro–Wilks tests confirmed the soil properties, gas production rates, and microbial abundance data were normally distributed within each population making them appropriate for analysis using parametric techniques. Soil properties and gas production rates for field cores and control samples were compared using a Student’s t test (n = 5 per group, df = 8). Effects of treatment (control, plant litter, or compost) on environmental variables, microbial abundance, EEA, and gas fluxes were analyzed using one-way analysis of variance (ANOVA; n = 5 per group, df = 14) with Tukey’s HSD for post hoc comparisons. Analyses were performed using the JMP statistical software (Version JMP Pro 9.0.2, Cary, NC, USA; Sall 2005) with a 0.05 significance level.

T-RFLP results were analyzed using principal coordinates analysis (PCoA) applied to the Bray–Curtis index of similarity derived from normalized fluorescence data; the first two axes from each analysis were plotted to visualize relative similarity in community structure across samples. Treatment effects were analyzed using a non-parametric multivariate ANOVA (NP-MANOVA), again applied to the Bray-Curtis similarity index. All community analyses were conducted using the PAST Version 2.16 statistical package (Hammer 2001).

For the 18-month data (July 2012), correlation analysis was performed to examine the relationships among the environmental, microbial, enzyme, and gas production data. (SPSS Statistics Version 20, Armonk, NY, USA). Multivariate normality was confirmed using Doornik and Hansen omnibus test in PAST prior to selecting Pearson’s coefficient.

Results

Controls versus field samples

After 6 months, the control samples did not differ significantly from field cores for any of the soil properties (pH, redox, OM, soil moisture, C:N; all |t| < 2.0 with p > 0.05). In contrast, for the 18-month sampling, t-tests revealed significant differences for all parameters except pH (for pH: t = 0.8 and p = 0.45; for all others, |t| > 2.5, all p < 0.05). Though the magnitude of the differences were small, redox (mV, mean ± SE; control: −128 ± 13, field: −83 ± 9), soil moisture (%; control: 52 ± 1, field: 60 ± 1), and OM (%; control: 7.8 ± 0.2, field: 9.7 ± 0.3) were all lower in the control samples; C:N was slightly higher (control: 9.6 ± 0.2, field: 8.4 ± 0.1). Gas flux rates (nmol g OM−1 h−1) were also measured for the 18-month sampling event, and no significant differences were observed for CO2 (control: 329.8 ± 50.4, field: 369.8 ± 120.0), CH4 (control: 5.1 ± 0.8, field: 40.9 ± 39.9), or total C gas production (control: 335.0 ± 50.9, field: 410.8 ± 157.5; all t < 1.0, p > 0.30).

Effects of organic matter manipulation

Environmental analyses

Addition of plant litter and compost increased soil OM relative to the controls; these differences persisted throughout the study (Fig. 1a; Table 2). After 6 months, OM content in the plant litter (14.5 %) and compost treatments (13.7 %) was similar to the levels at the start of the study (averaged across both treatments: 14.0 ± 1.4 %). However, after 18 months, average OM for these treatments decreased (litter: 10.0 %, compost: 11.9 %), but was still significantly higher than the unamended control (7.8 %).

Fig. 1
figure 1

Treatment effects on environmental parameters (a, b, c) and microbial abundance (d, e, f) following 6 (July 2011) and 18 (July 2012) months of in situ field incubation. Treatment effects were evaluated using one-way ANOVA within each sampling event; data are presented as mean ± SE, n = 5 per group. For all the graphs, lowercase letters denote statistically significant subgroups within the 6-month data set; letters with prime were used for the 18-month data. Treatments are color-coded and the symbols for the 6 month data are distinguished with a (plus) on the symbol

Full size image
Table 2 Statistical results evaluating treatment effects
Full size table

The compost and litter amendments also affected soil C:N, which was always lower in the controls (averaged by date: 9.5 ± 0.1) relative to the experimental treatments (Fig. 1b). At the 6-month sampling, average C:N was higher in the plant litter treatment (11.8) than in the compost treatment (10.6), reflecting a substantial decrease from the start of the study (Litter: 24.4 ± 2.4, Compost: 13.9 ± 0.2). These differences disappeared after 18 months of incubation (combined average for both treatments: 10.5 ± 0.2).

For all treatments at all times, redox potential was negative (Fig. 1c). After 6 months of incubation, there were modest differences between the plant litter (−54 mV) and compost treatments (−105 mV), but neither was significantly different from the control (−82 mV). After an additional year of incubation, there were no significant differences between any treatments (combined average across treatments: −135 ± 6.4), though values were generally more negative than at 6-months.

Soil moisture was consistently lower in the compost-amended soils (%; 6-months: 47.4 ± 0.5, 18-month: 50 ± 1.2) relative to the plant litter (6-months: 56 ± 1.1, 18-month: 52.1 ± 0.9) and control soils (6-months: 55.0 ± 2.1, 18-month: 57.1 ± 2.0) (Table 2). For pH, there were significant treatment effects at the 6-month sampling event only; pH was higher in the compost (6.1 ± 0.1) compared to the control (5.7 ± 0.1) or plant litter amendment (5.6 ± 0.1). There were no significance differences in porewater chemistry for any of the parameters (mg L−1, mean ± S.E.; DOC: 5.9 ± 0.9, TDN: 0.50 ± 0.08, TDP: 0.06 ± 0.01, NH4 +: 0.18 ± 0.04).

Microbial abundance and community structure

For all three groups, abundance was lowest in the compost-amended soils, where it changed little over time [gene copies g OM−1; averaged across both sampling events for bacteria: 17.9 × 109, archaea: 1.4 × 109, methanogens: 1.2 × 109 (Fig. 1 d–f)]. Bacterial abundance was ~fivefold higher in the control and litter-added soils for the 6-month sampling, and not significantly different at the 18-month sampling. For both times, archaea abundance in the control and litter-added soils was similar and ~ threefold higher than in the compost. For methanogens, significant differences were detected across all treatments for the 6-month sampling (gene copies g OM−1; control: 4.3 × 109, litter: 12.9 × 109, compost: 1.6 × 109). Abundance was slightly lower for all three treatments at the 18-month sampling, and the control and litter-added soils were no longer different (gene copies g OM−1; control and litter: 2.2 × 109, compost: 0.9 × 109).

Treatment effects on microbial community structure were visualized using PCoA (Fig. 2a–c) and statistical significance was evaluated using NP-MANOVA (Table 2). For all three groups and both sampling events, community structure in the plant litter treatment was significantly different from that in the control and the compost addition. Generally, the compost treatment did not significantly alter microbial community structure relative to the control; the only exception is for the archaea community at the 18-month sampling.

Fig. 2
figure 2

Treatment effects on microbial community structure following 6 (July 2011) and 18 (July 2012) months of in situ field incubation. Patterns in microbial community structure were visualized using PCoA and statistical significance was evaluated by NP-MANOVA. Data are presented as mean ± SE, n = 5 per group, and the percent of variance explained by each axis is provided. Statistically significant subgroups are designated using lowercase letters for the 6-month data set, and letters with prime were used for the 18-month data. Treatments are color-coded and the 6 month data are distinguished with a (plus) on the symbol

Full size image

Extracellular enzyme analysis

Treatment effects varied depending on the substrate tested and the kinetic parameter of interest (Table 2; Figs. 3 and 4; see Online Resource 1 for examples of the saturation curves from which these data were obtained). In general, the addition of plant litter corresponded to an increased Vmax relative to the control, which was statistically significant for BG (200 % higher), BX (550 %), and LAP (50 %). Conversely, the addition of compost suppressed Vmax relative to the control, though the trend was only statistically significant for AP (~25 % decrease). For Km, fewer treatment effects were observed, and there were never any differences in Km between the control and compost soils. In the presence of plant litter, Km was significantly lower for CBH (by 75 % vs. control) and LAP (by 15 %). The activity of POX was reduced by 10 % (relative to controls) in the litter-amended soils and by roughly 40 % when compost was added.

Fig. 3
figure 3

Effects of treatment on enzyme kinetics, evaluated after 18 months of in situ field incubation (mean ± 1 SE, n = 5 each; sampled July 2012): a maximum reaction velocity (Vmax) and b half-saturation constant (Km). Lowercase letters denote significant differences as determined via one-way ANOVA and Tukey’s HSD

Full size image
Fig. 4
figure 4

Phenol oxidase (POX) activity as affected by treatment, evaluated after 18 months of in situ field incubation (mean ± 1 SE, n = 5 each; sampled July 2012). Lowercase letters denote significant differences as determined via one-way ANOVA and Tukey’s HSD

Full size image

Anaerobic CO2 and CH4 production

Potential rates of CH4 and/or CO2 production in anaerobic slurries increased in response to the OM additions (Table 2; Fig. 5). The rate of CH4 production did not change significantly for compost but increased ~tenfold for plant litter. Rates of CO2 production in the compost and plant litter soils were 40–50 % higher than rates in the controls, with no significant differences in CO2 production between the two. Relative to the controls, total C gas production (CO2 + CH4) increased by 40 % in the compost treatment and by 70 % in the plant litter treatment (Fig. 5c). CH4 accounted for ~2 % of the total C gas production in the control and compost treatments, but that fraction increased to ~15 % in the treatment with the addition of plant litter (Fig. 5d).

Fig. 5
figure 5

Treatment effects on the production of CH4 (a), CO2 (b) and total C gas (c), as well as the fraction of total C gas that is CH4 (d), evaluated after 18 months of in situ field incubation measured in anaerobic slurries (mean ± 1 SE, n = 5 each; sampled July 2012). Lowercase letters denote significant differences as determined via one-way ANOVA and Tukey’s HSD

Full size image

Correlation analysis

When C gas production rates were correlated with the environmental variables, only the relationship between soil C:N and CO2 (r = 0.61, p = 0.01) was significant (other results not presented; all r < 0.50, p > 0.08). Similarly, microbial abundance did not show a strong relationship to gas production, except for the modest positive correlation of methanogen abundance and CH4 rates (Table 3). To examine how microbial community structure and gas production were linked, correlation analysis was performed using the scores from each PCoA. In each case, the first axis describing bacterial, archaeal, and methanogen community structure was strongly correlated with CH4 production (Table 3), and no significant correlations were obtained for any of the second axes (not presented; all r < 0.39 and p > 0.15). For EEA, CH4 production was significantly positively correlated with Vmax for BG, CBH, BX and LAP, and negatively correlated with Km for CBH and LAP (Table 3). The only significant correlation with CO2 production was for Km of BG (r = −0.60, p = 0.02).

Table 3 Pearson’s correlations coefficients (r, above and bolded) and p-values (below) associated with the comparison of gas production rates, microbial community attributes, and enzymatic variables at the July 2012 (18-month) sampling
Full size table

Several aspects of the microbial community were correlated with one another, including a significant relationship between the abundance of all three microbial groups (Table 3). Similarly, bacterial, archaeal, and methanogen community composition were correlated when each PCoA 1 was considered. Significant correlations were found between microbial community composition and EEA for Vmax for BG, BX and LAP, and with Km for CBH and LAP. In general, correlations were highest between EEA and bacterial or archaeal community composition, and were significant less often for methanogens.

Discussion

The loss of natural wetland ecosystems is often mitigated by construction or restoration of wetlands elsewhere in the watershed (EPA 2008). As a means of improving soil quality and promoting plant productivity, OM amendments are regularly included in these mitigation projects (Davis 1995; Mitsch and Gosselink 2000). Our study complements other research that has examined how this practice affects soil characteristics, redox gradients, vegetation, and nutrient cycling (O’Brien and Zedler 2006; Bruland et al. 2009; Sutton-Grier et al. 2009), and demonstrates that amendments can alter rates of C mineralization and induce shifts in microbial community structure and function. Furthermore, because our manipulation produced long-term changes in soil OM and C:N with relatively limited effects on other soil parameters (e.g., redox, pH, and soil moisture), we were able to isolate how OM characteristics can affect C biogeochemistry and identify key microbial drivers and feedbacks to the multi-stage process of decomposition. The similarities between measurements on control soils and intact field cores suggest that the results remain applicable to unaltered wetlands soils.

The amendments used in this study differ considerably in their biochemical composition. Compost, although derived from plant materials, undergoes a humification process that yields chemically-complex OM with few residual plant polymers and increased microbial necromass (Tiquia et al. 1996; Tuomela et al. 2000). Compost may be similar to native soil OM, which is also considered to have a significant portion of C of microbial origin (Simpson et al. 2007; Liang and Balser 2010; Throckmorton et al. 2012). In general, microbial necromass contains a larger fraction of proteins and lipids than does plant litter, and has only a small fraction of the carbohydrates and lignins dominant in plant litter (Nelson and Baldock 2005; Simpson et al. 2007; Throckmorton et al. 2012). Thus, while nutrient availability may have varied between the control and compost-amended soils, the chemical composition of the OM was probably more similar between these treatments and distinct from that in the litter-amended soil (that is, microbially-dominated vs. plant-dominated).

Treatment effects and carbon gas production

Overall, we found that both the plant litter and compost-added treatments exhibited higher potential rates of anaerobic C gas production relative to unamended soils on a per gram-OM basis (Fig. 5), which indicates that a greater fraction of the OM was mineralized compared to the control soil. One explanation for this result is that the OM in the plant litter and compost amendments was more labile than the existing soil OM. Additional factors at play include the potential for enhanced decomposition of native material through “priming” (Blagodatskaya and Kuzyakov 2008; Nottingham et al. 2009) or the possibility that some of the native soil OM was physically inaccessible (e.g., via sorption onto mineral surfaces), which could limit decomposition regardless of inherent lability (Kalbitz et al. 2000).

We did not identify any significant relationships between either soil environmental conditions or porewater chemistry and any of the following: microbial community composition, abundance, EEA, or C gas production, excepting a small positive correlation between C:N and CO2. Given the narrow range of C:N for our treatment soils (Fig. 1b), it is likely that C:N is not the driver of decomposition rates per se but instead a co-variant associated with finer-scale OM characteristics such as the degree of humification or OM lability as discussed above. The general lack of correlation between environmental variables and either the soil microbial community or biogeochemical response metrics suggests environmental conditions were not major divers of the observed treatment effects. It further supports our assertion that differences in C gas production were the result of microbial responses to OM type and not an unintended consequence of treatment on abiotic soil characteristics.

Role of microbial communities

Extracellular enzyme activity

Microorganisms can detect substrates in their environment and regulate enzyme production accordingly to balance resource needs with metabolic costs (Bhat and Bhat 1997; Shackle et al. 2000; Allison and Vitousek 2005; Allison et al. 2011; Shi 2011). In this study, we found considerable evidence that OM type (i.e., polymer availability) can influence EEA. For example, in the plant litter treatment, enzymes that target compounds abundant in plant litter (e.g., cellulose and hemicellulose) were elevated. This response was observed for Vmax of all of the C and N hydrolytic enzymes we measured, and was statistically significant for BG, BX, and LAP (Fig. 3; Table 2). Microbial adjustments of EEA are also evident in the compost treatment, where the reduced Vmax of AP was likely a response to decreased P limitation. The compost we used contained 0.2 % P, and microorganisms generally produce fewer acquisition enzymes for nutrients that are readily available (Sinsabaugh and Moorhead 1994; Allison and Vitousek 2005). Similarly, POX decreased approximately 50 % following compost addition, likely due to lower lignin content in compost (Sinsabaugh 2010).

Our results suggest that this variation in EEA across OM types may be due, at least in part, to changes in microbial community structure. Specifically, the decrease in Km associated with CBH and LAP in the plant litter treatment indicates the synthesis of isoenzymes with higher substrate affinity, which reflects more efficient allocation of resources (Marx et al. 2005; Stone et al. 2012). While multiple isoenzymes are known to occur within an individual organism (Esser et al. 2013), shifts in isoenzymes are also consistent with changes in microbial community composition (Farrell et al. 1994; Martinez et al. 1996; Tabatabai et al. 2002). The significant correlations between Km of CBH and LAP with bacteria PCoA 1 and archaea PCoA 1 (Table 3) further support our conceptual model that community structure can influence enzyme activity (Fig. 6). We anticipate that most of the EEA in our soils is of bacterial origin, as bacterial abundance was ~tenfold greater than that of archaea (average bacteria:archaea 16S rRNA gene ratio at the 18 month sampling, with a range from 4 to 21). Although enzyme parameters were correlated with archaea in our study (Table 3), and archaea have been demonstrated to produce extracellular enzymes in marine sediments (Lloyd et al. 2013), we suggest that these groups were not significant producers of enzymes in our system. This assertion is based on the fact that most of our archaea were likely methanogens (average archaea 16S rRNA:methanogen mcrA ratio was 1.3, with a range from 1.0 to 1.7), and methanogens exclusively use fermentation products and CO2 as their carbon source (Thauer et al. 2008; Reddy and De Laune 2008). Thus, it is unlikely they would expend resources to produce enzymes for carbon polymer breakdown (e.g., BG, CBH, BX, and POX) to liberate products they cannot directly utilize. Instead, we propose that the correlations we observed between EEA and archaea/methanogens are indirect based on methanogen consumption of fermentation products affecting upstream pathways of organic carbon breakdown (Fig. 6).

Fig. 6
figure 6

Conceptual model diagramming the hypothesized role of microbial community structure and extracellular enzyme activity in wetland organic matter decomposition. Microbially mediated flows of carbon are represented as thick arrows beginning with polymers and concluding with the terminal decomposition end products CO2 and CH4 (after Megonigal et al. 2004). Interactions between microbial structure and carbon pools/flows are designated as supported, not supported, or not tested in the current study

Full size image

Microbial community structure

We demonstrated that the plant litter amendment supported a distinct microbial community compared to the control and compost-added soils (Fig. 2a–c), and propose these differences developed in response to C substrate availability as mediated by EEA (Fig. 6). The initial mechanism for this OM effect is selection for a distinct set of heterotrophs capable of directly metabolizing the unique oligomers and monomers generated from EEA on plant litter. Given the current knowledge on wetland soil microbiology (Reddy and De Laune 2008; Wüst et al. 2009), we anticipate a large fraction of these organisms are fermentative bacteria. Fermentation generates acetate and other simple organic acids that support methanogens, the main archaea in our system. Thus we hypothesize the plant litter addition directly affected bacterial community structure (similar to Nemergut et al. 2010), and resulted in greater availability and altered composition of fermentation end products (e.g., acetate vs. propionate, Uz and Ogram 2006). Then, because many methanogen genera can use only a specific subset of fermentation products (Garcia et al. 2000), this altered substrate availability was the indirect mechanism for the observed change in community structure of methanogens (and archaea). These changes in community structure have the potential to impact C mineralization rates and the balance of CO2 and CH4 production.

Microbial regulation of C gas production

If polymer breakdown is the rate-limiting step in decomposition, there should be a positive correlation between EEA Vmax and C mineralization (e.g., Schimel and Weintraub 2003). In this study, no such relationships were observed for the CO2 production rates (Table 3). This may be partly due to the particular suite of enzymes we considered. Although commonly tested in soils, BG, BX, and CBH are fairly selective for plant polymers, and thus may not be as responsive to the availability of microbial necromass or humified material, which potentially dominated our control and compost treatments. Nonetheless, our results are consistent with the work of Freeman et al. (1997, 1998), who similarly found that BG activity did not correlate with CO2 production in wetland soils. The production of CO2 was also unrelated to microbial community structure, similar to the work of Bell et al. (2005) and Fromin et al. (2012). This may be because CO2 is generated by a plethora of microbial species with diverse metabolic strategies, creating considerable functional redundancy in natural communities (Botton et al. 2006; Griffiths et al. 2000).

In contrast to CO2, we did observe strong relationships between CH4 production, microbial community composition, and EEA, which we hypothesize are mediated through OM and bacterial community effects on the abundance, composition, and activity of methanogens (Fig. 6). Because methanogenesis is a fairly well-conserved function, performed by a monophyletic group of organisms (Garcia et al. 2000) that can utilize a limited range of organic substrates, there is relatively low functional redundancy associated with CH4 production and the contribution of individual species to overall ecosystem function should be more important (Allison and Martiny 2008; McGuire and Treseder 2010). We were able to identify two terminal restriction fragments (T-RF) in our data associated with the genus Methanosarcinales (after Smith et al. 2007) and found their relative abundance was positively correlated with CH4 production (Spearman correlation; T-RF 95 bp: r = 0.62, p = 0.01; T-RF 179 bp: r = 0.56, p = 0.02). These results suggest that specific taxa of methanogens may be strong drivers of CH4 production and are consistent with several other recent studies (Beckmann et al. 2011; Angel et al. 2012; Parkes et al. 2012). Additional research into the ecological and physiological attributes of these community members could further enhance our understanding of ecosystem-scale CH4 dynamics.

Conclusions

Our results have been used in conjunction with current knowledge on wetland decomposition to develop a conceptual model that incorporates microbial community structure and EEA to expand our understanding of CO2 and CH4 production rates (Fig. 6). Models such as this may be particularly helpful in understanding methanogenesis, since rates of CH4 production were strongly correlated with microbial community structure and multiple enzyme kinetic parameters. Relationships of enzyme activity and microbial community composition with CO2 production were considerably more tenuous. This may be a consequence of the numerous microorganisms, substrates, and metabolic pathways associated with anaerobic CO2 production (see Megonigal et al. 2004).

This work has direct implications for wetland restoration as plant litter and compost produced disparate changes in C gas production. Both OM sources increased total rates of anaerobic C mineralization relative to unamended soils, but only the plant litter additions increased rates of CH4 production (by roughly an order of magnitude). Similar results have also been reported for rice paddy soils (Singh et al. 2009; Ruirui et al. 2011), suggesting that the incorporation of highly decomposed OM amendments such as compost may help with wetland restoration (Stauffer and Brooks 1997; Sutton-Grier et al. 2009) while minimizing production of the greenhouse gas CH4.