Site History and Edaphic Features Override the Influence of Plant Species on Microbial Communities in Restored Tidal Freshwater Wetlands

Abstract

Restored wetland soils differ significantly in physical and chemical properties from their natural counterparts even when plant community compositions are similar, but effects of restoration on microbial community composition and function are not well understood. Here, we investigate plant-microbe relationships in restored and natural tidal freshwater wetlands from two subestuaries of the Chesapeake Bay. Soil samples were collected from the root zone of Typha latifolia, Phragmites australis, Peltandra virginica, and Lythrum salicaria. Soil microbial composition was assessed using 454 pyrosequencing, and genes representing bacteria, archaea, denitrification, methanogenesis, and methane oxidation were quantified. Our analysis revealed variation in some functional gene copy numbers between plant species within sites, but intersite comparisons did not reveal consistent plant-microbe trends. We observed more microbial variations between plant species in natural wetlands, where plants have been established for a long period of time. In the largest natural wetland site, sequences putatively matching methanogens accounted for ∼17% of all sequences, and the same wetland had the highest numbers of genes coding for methane coenzyme A reductase (mcrA). Sequences putatively matching aerobic methanotrophic bacteria and anaerobic methane-oxidizing archaea (ANME) were detected in all sites, suggesting that both aerobic and anaerobic methane oxidation are possible in these systems. Our data suggest that site history and edaphic features override the influence of plant species on microbial communities in restored wetlands.

INTRODUCTION

Diverse soil microbial communities, capable of using numerous metabolic processes to generate energy and assimilate nutrients, mediate key wetland functions. Although recent studies have described microbial community composition and functional gene abundances related to land use, vegetation, and environmental factors (1–3), structure-function relationships in freshwater wetland soils are not well understood. Biogeochemical activities are regulated not only by the size of the microbial biomass but also by the presence, distribution, and abundance of functional guilds (4). Therefore, functional gene markers can provide valuable insight into key biogeochemical processes and their relationships to site properties (5, 6). Given that the underlying mechanisms of major nutrient cycles are related to microbial taxonomic diversity, it is surprising that relatively few studies have described both microbial composition and functional group abundance in freshwater wetlands, a biogeochemical hot spot of carbon (C) and nitrogen (N) cycling.

Tidal freshwater wetlands (TFWs) are located in the upper reaches of estuaries along the coastlines of the U.S. Atlantic Ocean, the Gulf of Mexico, and elsewhere, where salinity is low (typically <0.5 ppt) (7–9). Unlike saline wetlands that tend to produce large quantities of hydrogen sulfide, the main C mineralization pathways in TFWs include methanogenesis (7, 8, 10) and, depending on mineralogy, iron reduction (11). The global contribution of methane from TFWs is unknown, but it is hypothesized to be negligible because of their limited area and competition with iron reduction (8, 11). However, the contrasting oxic and anoxic environments in TFWs support coupling of nitrification and denitrification, making these habitats important N sinks (12). Only a few studies have examined microbial community composition related to these processes in TFWs (10), and to our knowledge, no study has compared microbial composition between natural and restored TFWs.

Intense development in coastal zones has reduced TFW acreage and their associated ecosystem functions (7, 9). Efforts to restore these habitats unfortunately often fail to reinstate ecosystem services observed in natural wetlands, likely due to continued differences in abiotic and biotic factors (13, 14). Restoration of tidal wetland hydrology often necessitates lowering surface elevation by removing topsoil or raising it by depositing dredged sediment. These drastic alterations have direct impacts on physiochemical properties, such as bulk density, soil organic matter (SOM), and pH. Urban-impacted wetlands are particularly difficult to reestablish, because watershed development alters hydrology, nutrient flux, sedimentation patterns, and disturbance regimen, impacting the trajectory of plant community and soil development (9). It has become clear that restored wetland soils continue to differ from natural wetlands for decades or even centuries (13–16), but little is known about the effect of restoration on microbial communities and associated biogeochemical functioning in TFW (1, 17).

Wetland vegetation can impact soil microbes directly and indirectly. Microbial biomass and oxygen (O₂)-dependent metabolism are stimulated in the plant rhizosphere, where O₂ and C compounds are increased compared to levels in the surrounding soil (18–20). It has been observed that exotic plant species can significantly alter microbe-mediated function (21, 22). For example, soils under the Eurasian lineage of Phragmites australis had nitrification rates three times greater than that of the native Spartina patens in a brackish marsh (23, 24), and Lythrum salicaria tissue was observed to have a slower decomposition rate than the native Typha latifolia, leading to decreased nutrient pools (25, 26). However, other studies investigating plant-microbe dynamics, including in stands of Phragmites australis, reported negligible effects of plant species on microbial biomass C and N, soil respiration, denitrification, and potential net N mineralization (27, 28). These mixed results suggest mechanisms controlling microbial composition, and by extension, the processes they mediate are not well understood.

In the current study, we characterized bacterial and archaeal community composition and functional capacity via functional gene abundance in TFW soils from five locations, including natural and restored wetlands in urban and suburban watersheds. We hypothesized that soil properties such as SOM and mineral N concentration would differ between sites, and that these differences would correspond to differences in bacterial and archaeal composition and the abundance of functional genes. Furthermore, we tested if wetland microbial community composition and functional capacity would vary between plant species. For each of the five sites, we collected soil samples from the rhizosphere of four plant species: Typha latifolia (broad leaf cattail), Peltandra virginica (green arrow arum), Lythrum salicaria (purple loosestrife), and the Eurasian lineage of Phragmites australis (common reed). We examined the relative abundance of major phylogenetic groups and quantified 16S rRNA gene abundance for bacteria and archaea. In addition, quantitative PCR (qPCR) was used to measure functional genes representing denitrification (nirK, nirS, and nosZ), methanogenesis (mcrA), and methane oxidation (pmoA).

MATERIALS AND METHODS

Site description.

In July and August 2012, samples were collected from three restored and two natural reference TFWs. One natural (Jug Bay; lat 38.78580, long 76.71308; soil series, Nanticoke Mannington) and one restored (Wootons Landing; lat 38.85646, long 76.69124; soil series, Udorthents/water) site was located in the suburban area of central Maryland on the Patuxent subestuary of Chesapeake Bay. In 1992, soils were scraped down at Wootons to restore wetland hydrology (29). The lower Anacostia River is highly urbanized, as it enters Washington, DC, from central Maryland. In the Anacostia watershed, a natural remnant wetland (Dueling Creek; lat 38.92411, long W76.94018; soil series, Zekiah and Issue) was selected along with two restored marshes, one restored in 1992 to 1993 (Kenilworth; lat 38.91035, long 76.94588; no soil data available) and a second in 2000 (Kingman; lat 38.90414, long 76.96182; no soil data available). Kenilworth and Kingman sites were restored by raising the elevation with dredged Anacostia river sediments and then contoured with a Mud Cat dredge (30). Additional details for these three Anacostia sites are available in Baldwin (9).

Experimental design and sample collection.

For each of the sites, three replicated stands of four common plant species were targeted: Typha latifolia L., Peltandra virginica (L.) Schott, Lythrum salicaria L., and Phragmites australis (Cav.) Trin. ex Steud. subsp. australis. Each site contained areas dominated by these four species, with the exception of Lythrum, which was absent from Wootons. This study design resulted in a total of 57 collected samples. Aboveground biomass was clipped at the soil surface from a 625-cm² plot using a serrated knife and then placed in a large plastic bag to be separated later by species and dried to determine plant biomass (data not shown). After removing plant biomass, a half-circle Russian peat borer (Eijelkamp, Giesbeek, Netherlands) was used to collect two 5.2- by 50-cm cores. In each plot, soils were sampled <1 cm away from the clipped shoots of the species of interest. Cores were described in the field to identify major horizons (data not shown). The upper soil layers (i.e., the Oi horizons) were not observed in some restored locations; therefore, they were excluded from all samples. Remaining material from both cores was homogenized into a single representative sample and stored on ice until returning to the laboratory. Soil samples were thoroughly mixed, and ∼10 g of soil was removed from each sample and stored at −20°C until DNA extraction. The remaining soil was stored at 4°C until edaphic features were analyzed the following week.

Soil chemistry.

Soil pH was determined using an Accumet 15 plus pH meter on 5:1 water-soil slurries. Soil moisture content was determined by drying ∼10 g of field-moist soil to a constant mass at 105°C for 36 h. Soil organic matter was calculated using loss-on-ignition (400°C for 16 h) (31), and total C and N content was determined by combustion analysis at 950°C on a LECO CHN-2000 analyzer (LECO Corp., St. Joseph, MI) (32). Nitrate (micrograms NO₃⁻-N per gram dry soil) concentrations were determined by ion chromatography. Briefly, 5 g of soil was shaken in 12.5 ml of 0.1 M KCl for 1 h before centrifugation to pellet soil. The supernatant was passed through a 0.45-μm syringe filter to remove fine particles. The filtrate was stored at 4°C until analysis on an 850 professional IC autosampler (Metrohm USA, Inc., Riverview, FL) with an Metrosep A Supp 5-150/4.0 separation column and 20-μl injection. Ammonium (micrograms NH₄-N per gram dry soil) was extracted from 5 g of soil mixed with 2 M KCl and measured colorimetrically from the filtrate using a Multiskan FC spectrophotometer (Thermo Scientific, Waltham, MA) (33). Soil texture was determined using the hydrometer method (34) using composite samples from each site. Textures for each site were relatively similar: Jug Bay, from silt-loam to loam; Dueling, silt loam; Wootons, loam; Kenilworth and Kingman, both loamy sands.

Soil microbial characterization.

Total genomic DNA was extracted using a PowerSoil DNA isolation kit (Mo Bio Laboratories, Carlsbad, CA) by following the manufacturer's instructions, with the exception that soils were homogenized using a FastPrep-24 (45 s at 6 m/s; MP Biomedicals, LLC, Solon, OH). All samples were quantified using a Qubit 2.0 fluorometer (Life Technologies, Grand Island, NY).

Quantitative PCR.

Quantitative PCR was used to estimate the abundance of bacterial and archaeal 16S rRNA genes and seven functional genes: methyl coenzyme M reductase (mcrA), particulate methane monooxygenase (pmoA), ammonium monooxygenase α-subunit (amoA) for ammonia oxidizing archaea (AOA) and bacteria (AOB), nitric oxide reductase (nirk and nirS), and nitrous oxide reductase (nosZ).

Plasmid standards were constructed by amplifying functional genes from pure culture (see Table S1 in the supplemental material). Target genes of interest were amplified using a 20-μl PCR with the following reagent concentrations: 1× GoTaq colorless flexi buffer (Promega Corporation, Madison, WI), 1.75 mM MgCl₂, 0.20 mM deoxynucleoside triphosphates (dNTPs), 0.50 μM forward primer, 0.5 reverse primer, 0.064% bovine serum albumin (BSA), and 0.025 U/μl GoTaq hot-start polymerase (Promega Corporation, Madison, WI); details regarding primers, thermal cycling conditions, and efficiencies are listed in Table S1 in the supplemental material. Amplified functional gene fragments subsequently were cloned using the Topo TA cloning kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.

Prior to analysis, plasmid standards were linearized using EcoRV (Thermo Scientific, Waltham, MA) and purified using the UltraClean PCR cleanup kit (Mo Bio Laboratories, Carlsbad, CA). Standard plasmid concentrations were quantified using a Qubit 2.0 fluorometer (Life Technologies, Grand Island, NY) and subsequently adjusted to 2.5 ng/μl; this stock solution then was serially diluted 10-fold to 2.5 × 10⁻⁶ ng/μl. At least three of the six serially diluted standards were used to evaluate amplification efficiency and calculate gene copy numbers for the unknown environmental samples. Because reaction- and sample-specific inhibition can influence gene copy numbers, a soil standard dilution series was prepared to relativize plasmid curves (35). By following a procedure similar to that outlined in Hargreaves et al. (35), we prepared a soil standard by combining equal amounts of prediluted DNA samples. The pooled 2.5-ng/μl soil standard stock was serially diluted 10-fold to 2.5 × 10⁻⁶ ng/μl.

Soil DNA extracts, plasmid standards, and pooled soil standards were run in triplicate 20-μl reaction mixtures with 10.0 μl of KiCqStart SYBR green qPCR ReadyMix with ROX (Sigma, St. Louis, MO), 0.5 μM final concentration of each the forward and reverse primer, and 2.5 ng template DNA for community composition or 5 ng of template DNA for functional gene quantification. All reactions were run on the StepOne Plus real-time PCR instrument (Applied Biosystems, Foster City, CA).

Data were extracted from runs with standard curve r² values of >0.99, efficiency values between 90% and 110%, and a single dominant peak in dissociation curves (36). To calculate gene abundance for unknown samples, at least three of the six serially diluted plasmid standards were used to evaluate amplification efficiency. Additionally, threshold cycle (C_T) values were adjusted for differences between plasmid and soil standard efficiency according to equations outlined in Hargreaves et al. (35). Final gene abundance values (genes gram⁻¹ wet soil) were log transformed prior to statistical analysis.

Pyrosequencing.

Pyrosequencing was used to investigate microbial community structure. PCRs were set up using Promega GoTaq DNA polymerase (Promega Corporation, Madison, WI) by following the protocol described by Bates et al. (37). Each reaction was set up using primers F515 (5′-GTGCCAGCMGCCGCGGTAA-3′) and R806 (5′-GGACTACVSGGGTATCTAAT-3′), targeting a 291-bp fragment in the V4 and V5 region of 16S rRNA genes (37). This primer set was selected because it provides sufficient resolution for nearly all bacterial and archaeal organisms with few biases or excluded taxa (37). Multiplexing and sequencing of all 57 samples was accomplished using a 10-bp MIDS barcoded F515 primer also containing a Roche 454-A pyrosequencing adaptor (5′-CCATCTCATCCCTGCGTGTCTCCGACTCAG-3′; Roche Applied Science, Branford, CT, USA) and a GA linker sequence.

Target sequences were amplified in a 25-μl PCR. Each reaction mixture contained 0.20 μM forward and reverse primers, 0.20 mM dNTPs, 1.75 mM MgCl₂, 1× GoTaq colorless flexi buffer (Promega Corporation, Madison, WI) with 1.5 mM MgCl₂, 0.064% BSA, and 0.025 Taq U/μl GoTaq hot-start polymerase (Promega Corporation, Madison, WI). PCR conditions began with a 95°C heat activation step for 5 min, followed by 30 cycles of 95°C for 15 s, 55°C for 30 s, and 72°C for 30 s, with a final extension step at 72°C for 60 s. Postamplification, each barcoded PCR product was purified using an UltraClean PCR clean-up kit (Mo Bio Laboratories, Inc., Carlsbad, CA, USA), except 4.5× SpinBind solution was mixed with the 25-μl product. Separate sample amplifications were combined in equal amounts (37). The sample was sent to the Institute for Genome Sciences and Policy (Duke University, Durham, NC) and sequenced using titanium chemistry on a Roche 454 GS-FLX (Roche Applied Sciences, Penzberg, Germany).

Data analysis.

Prior to statistical analysis, each parameter was assessed for normality and homogeneity of variance assumptions. All variables except pH were log₁₀-transformed to meet normality assumptions. A split-plot design was analyzed using mixed-model analysis of variance (ANOVA) in the SAS System v. 9.2 (SAS Institute, Cary, NC) to evaluate the effects of site (whole-plot factor), plant species (subplot factor), and the plant-site interaction on soil parameters (pH, SOM, total C and N, NO₃-N, and NH₄-N) and microbial community functional genes (EUB, ARC, mcrA, pmoA, nirK, nirS, and nosZ) (38). The effects of plant species within each site (i.e., the simple effects) were included in ANOVA because of significant interaction terms for several of the dependent variables. Pearson's correlation coefficients were calculated between univariate data, and permutation tests were used to determine P values using Microsoft Excel.

Sequence data generated from the 454 sequencing runs were processed using the Quantitative Insights into Microbial Ecology (QIIME) pipeline (39). A full description of scripts and justification for each step is available (see Text S1 in the supplemental material). Briefly, sequences were demultiplexed and trimmed to remove barcodes, linker, and both forward and reverse primer base pairs. Sequences were quality checked using default settings in the split_libraries.py command, except minimum and maximum sequence length, and were adjusted to include the majority of sequences representing the 291-bp region. Samples were not denoised (40). Similar sequences were clustered into operational taxonomic units (OTUs) using Uclust, and a threshold with 97% similar sequence and taxonomy was assigned using the Greengenes database (http://greengenes.lbl.gov). The resulting relative abundances for each soil sample were used for subsequent analysis.

Rarefaction curves did not approach asymptote for all sample units (see Fig. S1 in the supplemental material). Due to unequal sampling depth among sample units, a rarified community was generated using the jackknifed_beta_diversity.py workflow script; rarefaction depth was set to the lowest sequence count (1,922 sequences). After rarifying the data set, unweighted principal component analysis (PCoA) was used. Recently it was reported that rarefaction removes valuable data and may lead to false conclusions (56); therefore, we also analyzed total community composition by site and plant species using the full quality-checked data set using nonmetric multidimensional scaling (NMS). NMS was performed in PC-ORD version 6 (MjM Software, Gleneden Beach, OR) to visualize overall differences in bacterial and archaeal 454 patterns across sites and plant species (41, 42). Analysis was performed using the Sorenson/Bray Curtis distance metric and random starting configurations with 250 runs with real data. Prior to analysis, rare species (less than 10 observations) were removed. A two-dimensional NMS with a final stress value of 9.7 was achieved after eight iterations and was used for subsequent analysis. The multiresponse permutation procedure (MRPP) was used to test for differences between sample units based on within-group similarities (42).

Sequence read accession number.

Sequences were submitted to GenBank as a single pooled sample under Sequence Read Archive accession number SRP055495.

RESULTS

Soil characteristics differed significantly among the five sites but varied little between the different plant species (Tables 1 to 3). Among the five sites, Jug Bay soils were more acidic and had higher concentrations of SOM, total C, total N, and NH₄-N (Table 1). The most recently restored site, Kingman, had less SOM, total C, total N, and NH₄-N than other locations. Dueling was more similar to the 1992 suburban restored site, Wootons, than to Jug Bay, its natural counterpart in the Patuxent subestuary. The site-plant interaction was significant for pH (Table 2) due to significant variation between plant species at the two natural sites, Jug Bay and Dueling (Table 3). Across sites and plant species, negative correlations were observed between pH values and SOM (r = −0.59, P < 0.01), total C (r = −0.56, P < 0.01), total N (r = −0.54, P < 0.01), and NH₄-N concentrations (r = −0.34, P = 0.01). Ammonium concentrations were positively correlated with SOM (r = 0.54, P < 0.01).

TABLE 1.

Soil characteristics for each of the five tidal freshwater wetland sites

Parameter	Value for each sitea
	Jug Bay	Dueling	Wootons	Kenilworth	Kingman
pH	4.6 ± 0.1	6.0 ± 0.2	6.0 ± 0.1	6.4 ± 0.1	6.3 ± 0.1
SOM	15.5 ± 2.3	6.1 ± 0.6	6.7 ± 0.4	5.0 ± 0.9	2.5 ± 0.6
Total C	7.8 ± 1.4	2.8 ± 0.4	2.9 ± 0.2	3.1 ± 0.4	1.3 ± 0.3
Total N	0.57 ± 0.10	0.19 ± 0.04	0.21 ± 0.01	0.18 ± 0.03	0.05 ± 0.02
NH4-N	16.2 ± 2.4	8.0 ± 0.9	9.8 ± 0.5	12.0 ± 2.6	5.9 ± 0.9
NO3-N	1.4 ± 0.1	1.6 ± 0.1	4.9 ± 1.9	1.3 ± 0.1	1.5 ± 0.1

^aResults are arithmetic means ± standard errors.

TABLE 3.

Results of ANOVA simple-effects tests of plant species within each site for soil characteristics and functional genes

Parameter	Value by sitea
	Jug Bay				Dueling				Wootons				Kenilworth				Kingman
	ndf	ddf	F	P value	ndf	ddf	F	P value	ndf	ddf	F	P value	ndf	ddf	F	P value	ndf	ddf	F	P value
Soil characteristics
pH	3	28.3	3.0	0.05*	3	28.3	5.0	<0.01*	2	28.3	1.0	0.39	3	28.3	1.7	0.19	3	28.3	0.9	0.45
SOM	3	38.0	0.9	0.44	3	38.0	0.3	0.86	2	38.0	0.1	0.88	3	38.0	1.1	0.35	3	38.0	2.1	0.12
TotalC	3	38.0	1.3	0.30	3	38.0	0.6	0.64	2	38.0	0.2	0.84	3	38.0	1.0	0.42	3	38.0	1.7	0.19
TotalN	3	38.0	0.8	0.50	3	38.0	0.6	0.61	2	38.0	0.1	0.91	3	38.0	0.9	0.44	3	38.0	2.5	0.07**
NH4-N	3	28.5	1.4	0.27	3	28.5	0.5	0.67	2	28.5	0.1	0.91	3	28.5	1.0	0.43	3	28.5	1.2	0.33
NO3-N	3	38.0	0.4	0.73	3	38.0	0.5	0.72	2	38.0	9.7	<0.01*	3	38.0	0.2	0.90	3	38.0	0.3	0.83
Functional genes
EUB	3	27.4	0.5	0.68	3	27.4	0.7	0.56	2	27.4	2.0	0.16	3	27.4	4.5	0.01*	3	28.4	3.3	0.04*
ARC	3	36.0	0.4	0.74	3	36.0	1.5	0.23	2	36.0	0.2	0.86	3	36.0	2.8	0.06**	3	36.0	2.6	0.07**
mcrA	3	27.1	0.6	0.65	3	27.1	1.5	0.24	2	27.1	1.1	0.34	3	27.1	0.1	0.96	3	28.1	0.3	0.84
pmoA	3	35.0	2.4	0.09**	3	35.0	0.8	0.50	2	35.0	0.4	0.65	3	35.0	1.7	0.18	3	35.0	0.6	0.60
nirK	3	27.6	1.7	0.19	3	27.6	14.3	<0.01*	2	27.6	3.6	0.04*	3	27.6	1.8	0.17	3	28.6	5.3	<0.01*
nirS	3	27.9	2.8	0.06**	3	27.9	1.0	0.42	2	27.9	1.0	0.39	3	27.9	2.3	0.10**	3	28.9	1.5	0.24
nosZ	3	36.0	1.2	0.32	3	36.0	1.4	0.27	2	36.0	3.8	0.03*	3	36.0	2.6	0.07**	3	36.0	3.4	0.03*

^aDegrees of freedom for the numerator (ndf) and denominator (ddf) were calculated using the Satterthwaite approximation. Significance is indicated as P ≤ 0.05 (*) or P ≤ 0.1 (**). F, F test statistic.

TABLE 2.

Results of ANOVA testing variation in soil characteristics and functional genes among sites, plant species, and the site-plant interaction^a

Parameter	Value forb:
	Site				Plant				Site-plant
	ndf	ddf	F	P value	ndf	ddf	F	P value	ndf	ddf	F	P value
Edaphic characteristics
pH	4	10.3	47.3	<0.01*	3	28.3	2.5	0.08**	11	28.3	2.4	0.03*
SOM	4	38.0	17.1	<0.01*	3	38.0	1.6	0.20	11	38.0	0.8	0.66
Total C	4	38.0	13.1	<0.01*	3	38.0	1.8	0.16	11	38.0	0.8	0.63
Total N	4	38.0	16.0	<0.01*	3	38.0	2.0	0.13	11	38.0	0.8	0.64
NH4-N	4	10.5	6.8	0.01*	3	28.5	1.7	0.19	11	28.5	0.7	0.76
NO3-N	4	38.0	10.4	<0.01*	3	38.0	2.2	0.11	11	38.0	1.6	0.16
Functional genes
EUB	4	10.4	3.7	0.04*	3	27.7	1.3	0.29	11	27.7	2.6	0.02*
ARC	4	36.0	13.5	<0.01*	3	36.0	1.0	0.40	11	36.0	1.8	0.08**
mcrA	4	10.1	3.8	0.04*	3	27.4	0.5	0.67	11	27.3	0.7	0.71
pmoA	4	35.0	6.0	<0.01*	3	35.0	0.9	0.43	11	35.0	1.4	0.23
nirK	4	10.5	15.4	<0.01*	3	27.8	4.3	0.01*	11	27.8	5.8	<0.01*
nirS	4	10.8	5.4	0.01*	3	28.2	0.7	0.56	11	28.1	2.1	0.06**
nosZ	4	36.0	7.1	<0.01*	3	36.0	0.4	0.76	11	36.0	2.9	0.01*

^aA SAS PROC mixed model was used to evaluate the whole-plot, completely randomized design.

^bDegrees of freedom for the numerator (ndf) and denominator (ddf) were calculated using the Satterthwaite approximation. Significance is indicated as P ≤ 0.05 (*) or P ≤ 0.1 (**). F, F test statistic.

Pyrosequencing generated a total of 478,143 quality-checked 16S rRNA sequences. Sequence lengths ranged between 166 and 317 bp, with the majority of sequences averaging 253 bp. Following the removal of low-quality sequences and chimeras, sequence counts ranged from 1,922 to 12,346 with an average sequence count of 8,388 per sample. When sequences were compared to the Greengenes database, 1,038 OTUs (97% similarity) were represented across all samples. Unclassified sequences were relatively few for all samples (means, 0.5%).

Ordination of the rarified sequence data set revealed microbial compositional differences between sites (Fig. 1a) but not by plant species (Fig. 1b). NMS ordinations and MRPP analysis of the nonrarified data resulted in a similar pattern, with microbial composition separated by site (see Fig. S2a in the supplemental material; P < 0.01 by MRPP) but not by plant species (see Fig. S2b). Microbial community composition correlated with pH (r = 0.49, P < 0.01) and with NO₃⁻-N (r = 0.52, P < 0.01) (see Fig. S2a). Most sequences putatively matched bacteria, averaging 79% of the total sequences per sample. The majority of the bacterial sequences were comprised of 12 phyla (Fig. 2a). Forty-eight to 72% of sequences with each sample matched one of these 12 phyla. The most abundant phylum across all samples was Proteobacteria (16%), with a large majority of sequences matching deltaproteobacterial (7%), betaproteobacterial (5%), alphaproteobacterial (2%), and gammaproteobacterial (1%) classes. Acidobacteria tended to make up a large percentage of abundance in Dueling (14%) and Wootons (13%) but accounted for only 6% of the relative sequence abundance in Jug Bay. The “Other” group in Fig. 2a refers to 58 additional phyla (12%) that were found in low abundance and unclassified bacterial sequences (5%). In general, relative proportions of bacteria to archaea were similar among all sites except in Jug Bay, where archaea made up a significant proportion of the microbial community (32%) (Fig. 2). The relative ratio of Euryarchaeota to Crenarchaeota was similar in all samples, and only a small percentage of sequences was unclassified archaea (0.5%; not plotted).

FIG 1.

Principal component analysis (PCoA) ordination of the microbial community composition rarified to 1,922 sequences per sample. Mean relative abundance ± standard errors (SE) is plotted by site alone (n = 12) (a) and site by plant species (n = 3) (b).

FIG 2.

Percent relative abundance of Bacteria (a) and Archaea (b) for five freshwater tidal wetlands (n = 12). The top 12 phyla in panel a represent the majority of the total bacterial sequences across all five sites (48 to 72%). The “Other” category represents the sum of 59 additional phyla, with 5% of the bar accounting for unclassified bacteria. The two major phyla in panel b represent 99% of the total identified archaeal sequences. Unclassified archaeal sequences are not shown.

Bacterial 16S rRNA gene copy numbers ranged from 2.3 × 10⁸ to 2.1 × 10¹⁰ genes g⁻¹ wet soil, with more bacterial gene copies in Wootons soils (1.1 × 10¹⁰ genes g⁻¹ wet soil) than Kingman soils (3.8 × 10⁹ genes g⁻¹ wet soil), with the other sites being intermediate (Fig. 3a). At both Kenilworth and Kingman, bacterial 16S rRNA gene copy numbers were lower in Peltandra than other plant genera, but this trend was not observed at other locations (Table 3 and Fig. 3a). Archaeal 16S rRNA ranged from 5.7 × 10⁶ to 2.2 × 10⁹ genes g⁻¹ wet soil and were significantly more abundant in Jug Bay (1.7 × 10⁹ genes g⁻¹ wet soil) than other locations (Table 2 and Fig. 3b). Similar to the bacterial 16S rRNA, plant species differences were observed (with Peltandra again having the lowest copy numbers) for archaeal gene copy numbers at Kenilworth and Kingman but were significant only at the 0.1 level (Table 2 and Fig. 3b). When the predicted ratio of archaea to bacteria using sequence data was plotted against the archaeal to bacterial 16S rRNA gene copy numbers, the ratios were significantly correlated (r = 0.92, P < 0.01) (see Fig. S3a in the supplemental material).

FIG 3.

Gene copy numbers gram⁻¹ of wet soil for genes targeting bacterial 16S rRNA (a), archaeal 16S rRNA (b), methyl coenzyme A reductase (mcrA) (c), and particulate methane monooxygenase (pmoA) (d). Values were calculated based on a linearized plasmid standard, and efficiencies were adjusted with a soil standard to account for inhibition. Each bar represents the mean (n = 3) ± SE. Note that panels have different y-axis ranges, and stars denote missing Lythrum at Wootons.

Copy numbers for some functional genes measured by qPCR differed between sites and plant species (Tables 2 and 3). Interactions between site and plant species were significant (some at the 0.1 level) for five of the seven genes examined, indicating that plant effects across sites were not uniform, but site effects were stronger than plant effects, based on lower P values for site than plant main effects (Table 2). Within individual sites, plant species related significantly to at least one function gene, with the exception of mcrA (Table 3 and Fig. 3c).

Minimal plant effects were observed for methanogens (Tables 2 and 3 and Fig. 3c), but gene copy numbers of mcrA were higher in Jug Bay (9.5 × 10⁸ genes g⁻¹ wet soil) than in the suburban reference site, Dueling (3.3 × 10⁸ genes g⁻¹ wet soil), and the three restored sites (Table 2). There was a positive correlation observed between mcrA and SOM (r = 0.35, P < 0.01). Examination of the methanogenic sequences revealed three classes of methanogenic Euryarchaea: Methanobacterium, Methanomicrobia, and Thermoplasmata. Eight families were represented in the sequence libraries: Methanobacteriaceae, Methanocellaceae, Methanomicrobiaceae, Methanoregulaceae, Methanospirillaceae, anaerobic methane-oxidizing archaea 2D (ANME-2D), Methanosaetaceae, and Methanosarcinaceae. Examination of the sequences found 93% of the sequences were dominated by four groups: Methanoplasmatales, Methanobacteriaceae, Methanoregulacae, and Methanosaetacae (Fig. 4a). The percentages of sequences putatively identified as methanogens were significantly correlated with the number of gene copies of mcrA (r = 0.46, P < 0.01) (see Fig. S3b in the supplemental material).

FIG 4.

Percent relative abundance of sequences putatively identified as belonging to methanogen (a) and methanotroph (b) taxa; bars represent the site means (n = 12) ± SE. Methanoplasmatales represents the recently reclassified Thermoplasmata (47).

Methanotroph pmoA gene abundance was greatest in Kenilworth (2.3 × 10⁵ genes g⁻¹ wet soil) and Wootons (1.6 × 10⁵ genes g⁻¹ wet soil) soils, followed by Dueling (9.1 × 10⁴ genes g⁻¹ wet soil), Jug Bay (6.4 ×10⁴ genes g⁻¹ wet soil), and Kingman (5.2 ×10⁴ genes g⁻¹ wet soil) (Table 2 and Fig. 3d). Sequences putatively identified as matching aerobic methanotrophs were present in all samples, including type I Gammaproteobacteria (order Methylococcales), type II Alphaproteobacteria (families Methylocystaceae and Methylobacteriaceae), NC10, and Verrucomicrobia (class Methylacidiphilae) (Fig. 4b). Similar to the methanogens, the percent abundance of sequences matching aerobic methanotrophs and pmoA gene copy numbers were significantly correlated (r = 0.33, P = 0.02) (see Fig. S3c in the supplemental material). Anaerobic methanotrophs (ANME-2D) also were detected in archaeal sequences across all five tidal freshwater wetland sites (Fig. 4b).

Ammonia-oxidizing bacterial genes were below the level of detection in all samples, and ammonia-oxidizing archaea genes were below detection limits in most samples; only 30% of the total samples fell within plasmid standard range (data not shown). In general, the effect of plant species on denitrification genes varied between sites (significant site-plant interactions [Table 2]), but some plant trends emerged. In Jug Bay, nirS gene copy numbers were higher under Phragmites than other plant genera (Table 3 and Fig. 5b; significant at the 0.1 level). The numbers of gene copies of nirK and nirS genes were lower in Jug Bay than other sites (Table 2 and Fig. 5) and were correlated with pH (r = 0.58, P < 0.01, and r = 0.56, P < 0.01, respectively) across all sites. Overall, gene copies of nitrous oxide reductase (nosZ) were highest in Wootons soils (Table 2 and Fig. 5c) and were correlated with SOM content (r = 0.45, P < 0.01), total C (r = 0.40, P < 0.01), total N (r = 0.49, P < 0.01), and NH₄-N (r = 0.39, P < 0.01) across all sites.

FIG 5.

Gene copy numbers gram⁻¹ of wet soil for genes targeting nitric oxide reductase (nirK) (a), nitric oxide reductase (nirS) (b), and nitrous oxide reductase (nosZ) (c). Values were calculated based on a linearized plasmid standard, and efficiencies were adjusted with a soil standard to account for inhibition. Each bar represents the mean (n = 3) ± SE. Stars denote missing Lythrum at Wootons.

DISCUSSION

Microbial community structure significantly differed between the five TFWs studied. Microbial community composition correlated with soil pH and NO₃⁻-N concentration (see Fig. S2 in the supplemental material). These findings partially support our hypothesis and corroborate other studies that have reported soil pH as an important factor shaping soil bacterial composition in many different ecosystems (43), including wetlands (1). Although this is a commonly reported finding, the mechanisms underlying these trends have not been fully explored. For example, Rousk et al. (44) and others presented evidence relating pH effects on microbial community composition; however, they did not find evidence for a link between different composition and C cycling functions (45). Interestingly, pH did significantly vary between plant species in the two natural sites (Table 3), suggesting that plants indirectly shape microbial communities in cases where vegetation has been established for a long period of time.

Significantly lower SOM content was observed in the urban and restored wetlands (Tables 1 and 2). Although we did not measure methane production, SOM content correlated with mcrA gene copy numbers (r = 0.35, P < 0.01), suggesting that there is increased potential for methane production in natural compared to restored sites. Putative hydrogenotrophic methanogens were dominant over acetoclastic sequences in all sites (Fig. 4a). This is in agreement with other studies of freshwater sediments, including peatlands and TFW sediments (10, 46). The most abundant group of methanogens matched a lineage of Thermoplasmatales that has been provisionally reclassified as Methanoplasmatales (47). These putative methylotrophic methanogens have been identified in many habitats, including another study of Jug Bay soils (10). Although there is not much known about this particular order, recent studies have shown that groups of methanogens vary in their O₂ sensitivity and available metabolic substrate (48–50). Seasonal O₂ penetration is relatively stable in TFWs (51); therefore, it could favor methanogen groups that are more sensitive to O₂. We plan to follow up this work by examining seasonal methane flux and tracking variations in the methanogen community through time.

We were surprised to find sequences putatively matching anaerobic methanotrophic archaea (ANME-2D) in all five of the wetlands. Anaerobic oxidation of methane (AOM) was first recognized in marine sediments and coupled with sulfate-reducing bacteria (52, 53), and we assume low levels of sulfate at all of our sites. However, recent studies have demonstrated the importance of AOM in TFW sediments and mudflats in situ (54). Furthermore, microcosm experiments demonstrated sulfate-independent AOM and coupled activity with alternative electron acceptors, including NO₃⁻, iron(III), and manganese(IV) (54). Although AOM sequences made up a higher relative abundance in the two natural reference sites, Jug Bay and Dueling, we documented relatively similar aerobic methane-oxidizing bacterial diversity (NC10, type I, type II, and Verrucomicrobia) (Fig. 4b). It is important to note that anaerobic methane-oxidizing archaea do not contain pmoA genes but instead contain mcrA; therefore, our qPCR targets do not clearly separate methanogenesis from methane oxidation. Given the abundance of ANME sequences, we plan to follow up this work to determine the relative contribution of aerobic and anaerobic methane oxidizers within TFWs and also to investigate the role of iron reduction as an alternative to methanogenesis.

Although we originally hypothesized that microbial communities would differ between the four plant species, we observed minimal difference in bacterial and archaeal community composition (Fig. 1b). Some functional gene copy numbers did vary between plant species within sites, but the effect of plant species was not uniform across site and tended to be weaker than site effects (Tables 2 and 3 and Fig. 3 and 5). Other studies have reported similar findings, concluding that edaphic properties and large landscape features may obscure plant-microbe relationships (27, 28). While we made an effort to sample the rhizosphere, the plant-affected area may comprise a small percentage of the soil, and our sampling efforts may have been too expansive to capture plant effects (20). Additionally, DNA analysis methods are limited and cannot capture dynamic changes due to radial oxygen leakage on microbial community composition or function. For example, denitrification genes are carried by numerous bacterial species, some of which may not express these genes if there is ample O₂ for aerobic respiration (55). Although we hypothesized that P. australis would support larger populations of aerobic functional groups, such as nitrifying archaea and bacteria, we found little evidence for amoA genes. These data suggest that even with radial oxygen leakage, the soils stay primarily anaerobic.

Conclusions.

Both restoration method and site legacy appear to be important factors affecting microbial community parameters. For example, we documented comparable compositions and functional gene abundances between Kenilworth and Kingman despite Kenilworth being restored 8 years earlier. The similar restoration methods used to restore Kenilworth and Kingman (use of dredged sediment as the substrate) may account for similar and persistent microbial communities. In contrast, composition in Wootons was significantly different, which may be attributed to the years of soil mining and the method of restoration (excavation to create tidal hydrology). Despite significant urbanization surrounding the Dueling site, microbial community composition was more similar to that of Jug Bay than to the three restored sites. We are encouraged that the small remnant wetland appears to maintain a microbial community similar to that of the suburban natural reference wetland, demonstrating the importance of conserving small TFWs in other urban centers. While plant metrics are commonly used as a proxy for wetland restoration success, our data suggest that differences in plant species, including native versus nonnative species, do not strongly affect microbial composition or functional potential, especially in restored wetlands. The main drivers of microbial composition and function appear to be related to substrate, surrounding land use, legacy, and restoration method.