Surface soil and soil cores were obtained from two sites (upper meadow, UM; lower meadow, LM) in the Pennington meadow, Baton Rouge, LA. The region experiences an average of 153.9 cm of precipitation annually (relatively evenly distributed) and an annual average temperature of 20.25⁰C. The Pennington meadow is about 8 hectares, and is bounded on the north and east by multi-lane city roads; a woodlot bounds the west, while Dawson Creek bounds the south (see Supplemental Figure 1). The site is not used actively, but is mowed for hay annually.

The UM site (30.400, −91.120) occurs at a topographic high along the upper margin of the watershed for Dawson Creek. This site rarely experiences flooding (none recorded), and precipitation percolates through the soils and flows down a shallow gradient toward the LM site (30.382, −91.128), which is adjacent to Dawson Creek. Because of its proximity to the creek, LM appears subject to fluctuations in groundwater levels as well as intermittent flooding. The LM site was approximately 30 m from Dawson Creek, while the UM site was approximately 0.7 km distant.

Two soil types occur in the meadow. The UM site is characterized by Oprairie series soils. These soils are classified as fine-silty, mixed, semiactive, thermic Fragiaquic Glossudalfs that are somewhat poorly drained and form in loess deposits on Pleistocene age terraces. The LM site is characterized by Deerfield-Verdun series soils. These soils are classified as fine-silty, mixed, superactive, thermic Albic Glossic Natraqualfs (Deerfield) or Glossic Natraqualfs (Verdun), which also form in Pleistocene age deposits.

During the August 2016 Baton Rouge Flood (see https://en.wikipedia.org/wiki/2016_Louisiana_floods for a general description), a large portion of the meadow was inundated by floodwater from Dawson Creek as indicated by flood maps and photographic evidence (Supplemental Figures 1, 2). However, due to its elevation UM was not flooded. The period of inundation for LM is unknown, but lasted for at least several days to 1 week based on anecdotal reports.

To assess the potential impacts of flooding, six surface soil samples (0–1 cm depth) were obtained from the LM site on 2 and 25 September 2016, and six were obtained from the UM site on 3 and 25 September 2016. Surface soil samples (0–1 cm) from both sites were collected at arbitrary points within an area of about 100 m². Soil was transferred to 100-ml Whirlpak sample bags that were stored at −80°C after collection.

In addition, six replicate intact cores (tubes 7.5 cm diameter x 30 cm length, soil about 10 cm depth) were collected from each site on 3 September 2016; the core bottoms were capped for transport to a laboratory at LSU where they maintained at ambient laboratory temperature (about 25°C) with an air headspace for the duration of the experimental manipulations.

Prior to experimental flooding, soil-atmosphere methane exchanges were assayed by sealing the cores to produce headspaces (about 700–800 cm³) that were sub-sampled (0.5 cm³) at intervals for immediate analysis with an SRI model 8610 gas chromatograph fitted with a flame ionization detector and standardized with a 10-ppm methane standard in hydrocarbon-free air prepared by diluting a SCOTTY 100-ppm methane in nitrogen certified standard (Sigma-Aldrich, St. Louis, MO). After terminating the gas exchange assays, three cores from each site were “flooded” with 120 ml of surface water obtained from Dawson Creek resulting in a water layer about 3 cm thick; the remaining cores were used as untreated controls. Flooded cores were not stirred during an incubation of 72 h.

Dawson Creek water for the flooding treatment was obtained on 3 September by filling a sterile 2-l bottle with water from the middle of the creek. The water was held at 5°C prior to use, and stirred with a magnetic stirrer. Coincident with addition to the soil cores, triplicate 20-ml sub-samples of Creek water were filtered through 0.2-μm pore size filters. The filters were stored at −80°C for subsequent extraction of genomic DNA (see below). As necessary, Dawson Creek water was added to the flooded cores (20–30 ml total) to maintain a 3-cm depth.

At the termination of the incubation, floodwater was removed from the cores using a pipet. A 20-ml sub-sample from each core was filtered as above. Cores were then sealed for assays of soil-atmosphere methane exchange as described above. Gas exchange incubations continued for approximately 40 h.

At the termination of the gas exchange assays, cores were extruded and sectioned into depth intervals of 0–1, 1–2, 2–4, 4–6, and 8–10 cm depths. Soil sub-samples from the interior of these depth intervals were transferred to Whirlpak bags, which were stored at −80°C until they were processed for genomic extractions. The remaining soil was sub-sampled for gravimetric water content, and stored at 4°C prior to drying at 80°C for 24 h. Additional sub-samples were used to measure pH. Briefly, soil samples were mixed with deionized water in a 1:2 ratio (mass of fresh soil:volume of water); pH of the mixture was assayed with a combination electrode and a Beckman Φ71 meter. Organic matter contents were assayed using a “loss on ignition” method (Davies, 1974). Briefly, dried soil samples from each of the core sections were weighed into porcelain crucibles and combusted at 550°C for a minimum of 8 h. The mass loss during combustion was considered an estimate or organic matter. Two cores (one each from flooded and non-flooded LM) were lost and not extracted.

Differences between sites, sampling dates, and flooding treatments, and among depths for pH, organic matter, and water content were assessed using the ANOVA module of XLSTAT (Addinsoft, Inc.). A Bonferroni correction was applied to the p-values for post-hoc comparisons. The independent variables (e.g., site, date, flooding treatment, and depth) were analyzed separately, and for interactions (e.g., site and depth) using a fixed effects model.

Genomic DNA was extracted from all soil samples (24 soil surface samples and 5 depths for each of 10 cores) using a PowerSoil kit (Qiagen Inc.) following the manufacturer's instructions. Genomic DNA was extracted from the 12 filters used for water sampling following the same procedure. After purification, DNA extracts were visualized by gel electrophoresis and then shipped overnight on dry ice to the Research Technology Support Facility of Michigan State University for 16S rRNA amplification (using primers 515f and 806r), barcoding, library preparation and sequencing on a Miseq platform (Illumina, Inc.) with 2 × 250 bp paired-end chemistry following the Kozich et al. (2013) protocol.

Sequences were analyzed using mothur (1.36) following the Miseq protocol (Kozich et al., 2013) to make contigs, trim, filter, align, remove chimeras, classify and cluster OTUs (d = 0.03). The SILVA v 128 database was used for both alignment and classification (Quast et al., 2013). Alpha diversity metrics for each sample group, including Chao index, ACE index, Shannon index, inverse Simpson's index and coverage, were estimated after removing singletons and normalizing the data set. Beta diversity metrics indicating shared diversity were based on Yue-Clayton theta (Yue and Clayton, 2005) and Jaccard (1912) dissimilarities. Pairwise distances were used to generate a dendrogram representing relationships among samples. Guided by these outcomes, additional analyses were conducted using the Microbiomeanalyst platform (Dhariwal et al., 2017). The most informative OTUs were selected using a filtering protocol based on mean abundances for OTU counts, and the inter-quantile range function for OTU variance. Data were transformed using the centered log ratio, but samples were not rarefied because the variation in library sizes was small. Analyses included standard estimates of alpha diversity (Chao1 [richness] and Shannon indices [abundance and evenness]) and beta diversity (Bray-Curtis plus unweighted and weighted UniFrac distance metrics) with ordination by principle coordinates analyses (PCoA) with permutational analysis of variance (Permanova) and permuted homogeneity of multivariate dispersions (Permdisp), plus Linear Discriminant Analysis Effect Size (LEfSe) to assess features (OTUs) that served as biomarkers for the various sample groups. Sequences have been deposited in the NCBI SRA as SRP176467.

Soil pH varied between 4.73 and 5.69 for upper meadow cores, and between 5.28 and 5.89 for lower meadow cores (Supplemental Figure 3); the pH of Dawson Creek water samples was somewhat higher, 6.60–6.72. Analysis of variance (ANOVA) of soil pH values with a Bonferroni post hoc test (critical value = 0.008) (Supplemental Table 2) revealed no statistically significant impact of experimental flooding (p > 0.215), although flooded soils tended to have lower pH values. Experimental flooding in LM cores was associated with higher pH values than in the unflooded cores, but the differences were only marginally significant (p = 0.09). However, the pH of the experimentally flooded LM soils was significantly higher than both sets of UM cores (p < 0.0001).

Organic contents were greatest in the surface interval, and decreased rapidly with increasing depth (Supplemental Figure 4). For the upper 1 cm, values varied between 15.4 and 20.1% for UM cores, and 29.1 and 21.4% for LM cores. Values at 9 cm were <4.3% for both sites and treatments. Organic matter contents were otherwise similar among sites and treatments (p > 0.210) (Supplemental Table 2).

Water contents varied significantly with increasing depth for soil cores (p < 0.0001, Figure 1), but did not vary between LM and UM sites when experimentally flooded or unflooded cores were compared separately. However, flooding for 3 days at both sites significantly increased water contents over the 0–4 cm depth interval relative to unflooded soils (p < 0.0001, Figure 1), while lower depths were unaffected (Supplemental Table 2).

Soil-atmosphere methane fluxes differed distinctly between LM and UM, and responded differentially to experimental flooding at the two sites. In particular, LM cores prior to flooding consistently emitted methane (Figure 2; 1.3 ± 0.8 mg m⁻² d⁻¹ [mean ± 1 standard error of the mean], n = 3), while UM cores consumed it to values less than ambient atmospheric concentrations (Figure 2; −0.3 ± 0.2 mg m⁻² d⁻¹, n = 3). Subsequent to flooding, methane fluxes in LM cores increased relative to pre-flooding values (Figure 2; 6.2 ± 0.6 mg m⁻² d⁻¹, n = 3), while unflooded cores continued to emit methane, but at rates lower than for flooded cores (Figure 2; 1.5 ± 1.2 mg m⁻² d⁻¹, n = 3). UM cores that previously consumed atmospheric methane emitted it after flooding (Figure 2; 1.6 ± 0.7 mg m⁻² d⁻¹, n = 3), while consumption continued for unflooded cores (Figure 2; −0.8 ± 0.5 mg m⁻² d⁻¹, n = 3).

Genomic DNA was successfully extracted from 74 soil samples. Fifty of these were from 10 cores, each sampled at 5 depths; two of the initial 12 cores were lost during processing, one each from the LM flooded and unflooded sets. An additional 24 surface samples were collected from the LM and UM sites on 2 and 25 September, 2016. Of the 74 soil extracts, sequence libraries for two were excluded from the analysis pipeline due to low read yields. In addition, genomic DNA was extracted from 12 Dawson Creek water samples, including 3 that were filtered immediately after water collection, 3 that were stored at 15°C during the experimental period, and 3 each from the LM and UM cores that were flooded. A total of 7,140,700 reads were obtained from the Miseq run. Of these, 3,765,465 passed quality and filtering screens, and were assembled into contigs (mean library size = 45,367 reads, standard deviation = 9,546). After eliminating singletons, 64,525 OTUs defined at an evolutionary distance of 0.03 remained from mothur's cluster split algorithm.

Irrespective of sample location, date or treatment, four phyla dominated all communities. Acidobacteria and Proteobacteria were most abundant, followed by Verrucomicrobia and Bacteroidetes. These phyla collectively accounted for 74.7%-89.9% of all OTUs. Relative abundances varied somewhat among soils as a function of location, sampling date and depth, with ranges of 21.0–28.7%, 24.1–32.0%, 14.1–20.1%, and 4.1–11.5% for Acidobacteria, Proteobacteria, Verrucomicrobia, and Bacteroidetes, respectively (Supplemental Table 1). For Dawson Creek and the experimental floodwater samples, Proteobacteria and Bacteroidetes were generally more abundant than in soils, while Acidobacteria and Verrucomicrobia were lower (Supplemental Table 1).

Numerous phyla occurred at abundances <1%, including Actinobacteria, Armatamonadetes, Chloroflexi, Firmicutes, Gemmatimonadetes, Planctomycetes, and Spirochaetes (Supplemental Table 1). Various candidate divisions and phyla were also present at low abundances, e.g., BRC1, Microgenomates, Latescibacteria, Parcubacteria, Saccharibacteria (formerly TM7), and WPS-1, WPS-2 (Supplemental Table 1). No distinct patterns for distribution of these phyla were evident.

The Acidobacteria were notable for the numerous phylogenetic groups that were represented in all samples (Figure 3). Groups 1, 3, 4, and 6 were particularly abundant, but 2, 5, 7, and 10 were also common, while Groups 11, 13, 15–18, 22, and 25 were present at low abundances (<0.1% overall). Alphaproteobacteria were generally more abundant in soil samples than Delta- and Betaproteobacteria, which were comparable in abundance, and substantially greater than Gammaproteobacteria; Epsiloproteobacteria were rare (Figure 3). In Dawson Creek and experimental floodwater samples, Betaproteobacteria were typically more abundant than other Proteobacteria classes (Figure 3). Among the Verrucomicrobia, the genus Opitutus was relatively abundant (0.6% overall), but unclassified OTUs belonging to Spartobacteria and Subdivision 3 represented large fractions of all libraries (7.3 and 5.3% overall, respectively), and dominated the Verrucomicrobia (Figure 3).

Numerous OTUs were classified in functionally distinct families or genera. Methanotrophs were represented in the Alpha- and Gammaproteobacteria, but at a low incidence (0.04% each overall) with no distinct patterns among samples. OTUs classified as methanogenic Euryarchaeota were also present at a low abundance in soil (0.04% overall) but not in water samples. OTUs associated with nitrification were dominated by Nitrospirae (0.44% overall) and Nitrososphaera (Thaumarchaeota, 0.25% overall); proteobacterial nitrifiers (e.g., Nitrobacter, Nitrosomonas) were rare. In contrast, proteobacterial nitrogen fixers classified as members of the Bradyrhizobiaceae were relatively abundant (1.09% overall), though nitrogen fixers classified as Rhizobiaceae were considerably less common (0.06% overall).

The 20 most abundant OTUs in the dataset collectively accounted for 21.9% of all of the reads (Table 1). They included representatives from Acidobacteria, Actinobacteria, Bacteroidetes, Proteobacteria, Verrucomicrobia, and unclassified Bacteria. Two unclassified Spartobacteria (Verrucomicrobia) were the most abundant OTUs overall. Together with a representative of Verrucomicrobia subdivision 3, they accounted for 7.26% of total abundance. Acidobacteria (8 OTUs in the top 20) were represented by OTUs classified in Groups 1, 3, 4, and 6, collectively accounting for 7.47% of the total abundance (Table 1). Proteobacteria, the third most abundant phylum among the top 20 OTUs, was represented by 6 OTUs from the Alpha-, Beta-, and Deltaproteobacteria that collectively accounted for 3.39% of the total abundance. While many of the OTUs in this study could not be classified below a family level, four genera were found among the top 20 OTUs. They included Terrimonas (Bacteroidetes), Bradyrhizobium (Alphaproteobacteria), and Actinoallomurus (Actinobacteria), which collectively accounted for 2.78% of all reads, while 2 Geobacter OTUs (Deltaproteobacteria) accounted for an additional 1.56%.

Alpha and beta diversity indices were generated for the various samples using mothur after removing singletons and normalizing library abundances. Chao1 estimates of richness ranged from a minimum of 5,213 ± 1,556 for stored Dawson Creek samples to a maximum of 10,526 ± 747 for values integrated over the UM experimentally flooded cores (Table 2). However, there were no statistically significant differences in Chao1 among sample sources, dates or treatments, even though water samples consistently harbored lower richness, and UM soils were richer than LM soils. Similarly, no statistically significant trends were observed for Chao1 as a function of core depth.

Inverse Simpsons and Shannon diversity indices also did not vary significantly among sample sources, dates or treatments. The former varied from a minimum of 96.8 ± 24.7 to a maximum of 235.2 ± 23.2 in LM experimentally flooded cores and 2 September 2016 LM surface soils, respectively (Table 2). The distribution of Shannon diversity values largely paralleled that for the inverse Simpsons index, but with the highest value, 6.945 ± 0.044, observed for 2 September 2016 UM surface soils. Both indices tended to decrease with increasing soil depth, and both were higher in UM than LM soils. However, neither trend was statistically significant.

Relationships among soil microbial communities were initially explored with the Yue-Clayton theta similarity index (Yue and Clayton, 2005) as implemented in mothur with sample groups clustered using a UPGMA algorithm (Supplemental Figure 5). The initial results indicated that LM and UM soils were distinct, but that the flooding treatment had no notable impact on microbial community composition at either site. The results also indicated that communities in surface soils collected on 2 September 2016 differed from those collected on 25 September 2016, especially for LM. In addition, the results revealed that the composition of Dawson Creek samples differed from all of the soil samples. However, two of three water samples collected from each set of the flooded cores clustered with soil samples rather than Dawson Creek samples (Supplemental Figure 5).

Based on the initial findings, principal coordinates analyses (PCoA) were used to explore relationships among communities based on sample date (2 and 25 September 2016 surface samples) and location (LM vs. UM) using Bray-Curtis, unweighted UniFrac, and weighted UniFrac beta diversity indices based on filtered data sets with the most informative OTUs transformed using a centered log ratio algorithm as implemented by Microbiomeanalyst. The results were largely comparable to those obtained using mothur's tree.shared command and the thetaYC index (Supplemental Figure 5). PCoA analyses for LM and UM surface soils collected on 2 and 25 September 2016 were based on 1,827 significant features. Results showed that surface soils from the two sites were distinct, and that communities from 2 and 25 September differed for LM, but not UM (Figure 4; unweighted unifrac, Permanova: R² = 0.452; p < 0.001; Permdisp: p = 0.467). PCoA results also showed that LM and UM communities as represented by experimentally flooded and unflooded cores at all depths (1928 significant features) were distinct, e.g., Figure 5 (Permanova: R² = 0.187, p < 0.001; Permdisp: p = 0.002), with the first two axes accounting for 34.3% of community variability. However, experimental flooding did not alter communities in either location (Supplemental Figure 5, Figure 6).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.